You are on page 1of 117

The Issues of Competition in Mainframe and Associated Services in India

Rajat Kathuria Subhasis Bera Shikha Gupta Mamta Raj Kumar Shahi

1

Contents Abstract ..................................................................................................................................... i Executive Summary ................................................................................................................. ii I. Introduction ......................................................................................................................... 1 II. Developing a Framework for Analysis .............................................................................. 2 III. Coming to grips with the relevant market from the Supply and Demand Side ................ 7 IV. Indian Server Market: Prospects....................................................................................... 9 V. Structure of the Server Market in India and APAC ......................................................... 13 VI. Primary Survey: Results ................................................................................................. 20 VII: Analysis of Competition in the High End Server Market in India: Is it Adequate? ..... 29 VIII. Abuse of Dominance in the Presence of Network Effects and High Switching Costs 32 IX. Conclusion: Future Market Development, Competition Policy and role of CCI ........... 38 Bibliography .......................................................................................................................... 49 Annex I................................................................................................................................... 54 Annex II ................................................................................................................................. 55 Annex III ................................................................................................................................ 56 Annex IV ................................................................................................................................ 57 Annex V ................................................................................................................................. 58 Annex VI................................................................................................................................ 63 Annex VII .............................................................................................................................. 64 Annex VIII ............................................................................................................................. 66 Annex IX ................................................................................................................................ 68 Annex X ................................................................................................................................. 71 Annex XI................................................................................................................................ 80 Annex XII .............................................................................................................................. 87 Annex XIII ............................................................................................................................. 88 Annex XIV ............................................................................................................................. 89 Annex XV .............................................................................................................................. 93 Annex XVI ............................................................................................................................. 94 Annex XVII ........................................................................................................................... 95 Annex XVIII .......................................................................................................................... 96 Annex XIX ............................................................................................................................. 98 Annex XX ............................................................................................................................ 100 Annex XXI ........................................................................................................................... 101 Annex XXII ......................................................................................................................... 102 Acknowledgements .............................................................................................................. 103

List of Tables Table 1: Table 2: Table 3: Table 4: Table 5: Table 6: Table 7: Taxonomy of the market based on selling Price ......................................................... 8 Total ICT Expenditure US $ Million, Asia Pacific .................................................. 10 Revenue and Shipment growth for Asia Pacific ....................................................... 12 Share of Mainframe in IBMs Total Revenue for India, China and APAC............... 14 HHI for Asia Pacific (APAC) by Vendor Revenue .................................................. 15 HHI for Asia Pacific (APAC) by Shipment ............................................................. 15 Percentage Share of Operating System by Vendor in the High End Market by Shipment .................................................................................................................... 20 Table 8: Extent of Competition in the three segments ............................................................ 22 Table 9: Intensity of competition in the submarkets............................................................... 23 Table 10: Mainframe Clients added in 2008 in India ............................................................. 24 Table 11: Vendors providing servers to the organization ....................................................... 26 Table 12: Main uses of Server infrastructure.......................................................................... 26 Table 13: Server architecture organization is predominantly using ....................................... 27 Table 14: Operating Systems running on the servers ............................................................. 27 Table 15: Frequency of upgrading / replacement of the existing server infrastructure .......... 28 Table 16: Percentage of server infrastructure used to run mission-critical data ..................... 29 Table 17: Market Shares in the High-end Segment ................................................................ 30 Table 18: R&D and Advertising for the Big Three ................................................................ 31 Table 19: Sources of Market Power for Microsoft and IBM Compared ................................ 36 Table 20: Sector-Specific Policy for FDI in India .................................................................. 39 Table 21: Mainframe and Superdome’s average price in India and China............................. 42

At the same time the report cautions that expansion in the installed base of mainframes with the proprietary z/OS could lead to welfare losses like those reported for Europe and suggests a possible role for the Competition Commission of India (CCI) under the existing legal framework. L1. Regulation i . These reveal that the market is highly concentrated. Since structural indicators of competition do not always reflect or imply abuse. Monopoly. This is the first study to anlayse competition and related issues in the Indian server market. especially in the high end segment. While competition to the mainframe has developed in recent times. with an extensive focus on mainframe computing. Antitrust Law. our survey also points to the difficulty of migrating away from a proprietary technology such as the z/OS owned by IBM and tied to its mainframe hardware. The survey of users across different size classes and verticals belonging to both the private and public sectors reveals that by being late starters. Indian users were able to avoid many of the costs associated with being locked in to a proprietary technology such as the mainframe. Concentration. Structural indicators of competition are estimated using secondary data across different segments of the server market. ___________________________________ JEL Classification: D4. Anti Competitive. firm conduct is gauged from an extensive primary survey of users and vendors. L40 KEY WORDS: Market Structure. L20. K21. Competition. The low installed base of such systems in India compared to Europe and the United States implies that there are no immediate public policy concerns in this regard.Abstract Very little is known about the extent and nature of competition in the mainframe and associated services market in India. Concentration progressively reduces in the mid and entry level segments respectively.

At the same time rivals of IBM. have also come up with a wide range of mainframe like products to provide more competition in the high end server market. at current prices < 25000 Entry level servers also known as volume server market >25. we use the first two definitions. 4. by server selling price. If the Herfindahl is low. The levels of concentration reduce slightly if we use shipment data instead of revenue. which does not yet market these computers in India.Executive Summary 1. there are many competitors and exercising market power should be hard. Albeit rapid technological development was expected to ring the death knell for mainframe computing as smaller computers became more powerful and microprocessors more multipurpose.keeping the objective of the study in mind. We do this for both revenue and shipment data for the APAC region as a whole and for individual countries. The structural estimates of competition are measured using Herfindahl-Hirschman Index (HHI). We use the following price bands to classify the market. IBMs revenue from mainframe in India is minuscule compared to its overall revenue. a high Herfindahl. although the qualitative result does not change i. This study focuses on understanding the nature and extent of competition in the server market in India by investigating both structure and conduct. Prices (US $). It is noteworthy that the high end market is effectively a triopoly between IBM. including mainframes 3.000 Midrange enterprise server market >100. The numbers show that the high end server market on average is more concentrated than the mid range market. Global IT market research firms such as Gartner and IDC also classify the server market by average selling price and therefore this taxonomy makes it easier to compare our findings for India with the findings for the broader Asia market and the world market as a whole. with US$ 100. on the other hand. by types of processor. Definition of the server market is a thorny issue as there is no universally acceptable standard. The levels of concentration over time have not altered much either. by server operating system.e.000 as the cut off point for the high end configuration based on our finding that the Indian market is relatively more ‘price sensitive’. which is determined by adding the squares of the market shares of all firms and C4 (market share of the top 4 firms in the segment). 2. In fact the only other manufacturer of mainframes in the world today is Fujitsu. which in turn is more concentrated than the entry level or volume market. Although there are at least five traditional ways of segmenting this marketby server hardware brand. over the years the mainframe has become even more powerful and versatile. on the other hand it is much higher in ii . HP and Sun Microsystems with the latter a distant third.000-100. by number of processor sockets in the market. the only surviving mainframe vendor. results in a concentrated market in which price rises may be easier to sustain. In other words.000 High-end enterprise server market. this segment is highly concentrated with the 4 firm concentration ratio measuring 100 for the entire sample period. We are unable to calculate structural estimates of market concentration for only the ‘mainframe’ segment since there is no competitor to the IBM mainframe in India. the high end and medium range markets are persistently more concentrated than the low or volume market.

Across the three segments that we are studying. performance of x86 processors has improved manifold over the last 10 years providing benefits of “value and standardization”. not only for running financial market transactions but also to support millions of lines of programming code that are being developed for the rest of the mainframes circling the globe. in fact Dell’s entire range of servers is sold with the x86 range. Customers exercise choice over a specific type of processor and OS (performance of OS depends upon the number of processors in a multiprocessor environment and performance of application software largely depends on the choice of OS). Accordingly we distinguish between the ex ante nature of competition and ex post competition within this segment based on whether the application is a new workload capable of being handled by any server in the high end segment or a legacy workload. In other words. the x86 processor dominates the entry level market across all vendors.83 trillion from domestic business in India for fiscal 2009. according to industry experts. up from Rs 42. the microprocessor and the Operating System (OS) are often bundled with the hardware by vendors. however Sun Microsystems market share in India at 15% presently is quite low. database management software (industry experience says performance of database varies across the sectors). This implies the average price of x86 processors is lower than other processors. About 90 per cent of all IBM servers sold in India in 2009 included x86 processors. 6. Mainframe installations and skills are growing in India. For example. Only Sun Microsystems sold more servers with RISC processor than with x86. Although we club the mainframe market with the high end for the sake of measuring concentration. by clubbing multiple small servers one can get performance akin to high end computing. IBMs India-specific revenue is estimated to be over $1 billion. In the early 90’s there were no mainframes in India or China. At the entry level. This number progressively decreases to about 40 per cent and 0 for the mid and high end market segments respectively. one high end server can replace number of entry level servers. a customer can choose a single vendor or opt for a multivendor arrangement. We therefore limit our analysis to the choice of OS and choice of processor which will give us an indication of the ‘revealed preferences’ in the server market in India. 90 per cent of IBMs systems come bundled with x86 processors in the entry level market. for which the options are rather limited. The preferred processor for the entry level market is the x86 processor. it may not be appropriate to combine the high end market and the mainframe market.80 trillion in fiscal 2007.42 trillion in fiscal 2008 and Rs 33. Depending on these needs. According to Intel. X86 processors offer flexibility to the customer in the choice of OS and applications. dominance of x86 processors can be attributed to the customer’s choice for open technology. the qualitative analysis of competition and assessment of conduct is carried out recognising that under certain conditions. This indicates that a iii . application software and other technical specificities. 7. The two key building blocks of servers.China and APAC. Although it is important to analyse each preference type. For example the ‘lock in’ for legacy applications is likely to be different in the high end market compared to the mainframe market. IBM had revenue of Rs 57. 5. it is nearly impossible to track all the information for each firm across each vertical. X86 is an open technology architecture that enables wider choices for the client. The same analysis on the basis of revenue rather than shipments shows that the share of x86 in overall revenue is about 55 per cent. In addition.

Linux deployment has been increasing in the last 3 years and captured 20 per cent market share in terms of shipments and 11 per cent market share in terms of revenue in 2008. For legacy workloads this will be true only if the applications and operating system running on the high-end server are compatible with the small servers in the cluster. HP and IBM versions of Unix that run on their high-end servers are not compatible with cheaper x86-x64 servers that are run on Linux or Windows Server. while UNIX dominates the mid to high level segments. choice of OS is also likely to vary across verticals. Thus for new workloads the Superdome’s hardware performance can match that of IBM mainframes. The market share for the IA64 Itanium processor used in HP machines has increased while that of the RISC processor in HP has declined since HP stopped selling servers based on its RISC chip in 2008. service providers and a representative cohort of medium and large enterprises from different verticals. However. 10. for example HCL has its manufacturing capabilities in Chennai and Pondicherry. where these legacy mainframe applications are concerned. although users of IBM mainframe operating system and their associated legacy workloads are pretty much locked in. Our survey reveals that choice of OS is dictated by the application software the client wants to run. 9. In India activities of server vendors are not region specific and almost all servers sold locally by MNC vendors are imported from other manufacturing locations which implies that they are covered by the import duty regime. since the Superdome cannot support these latter types of software. Some large MNC vendors leverage their desktop/notebook capabilities to produce the low-end volume servers. Across segments the results are predictable. Our primary sample (purposive and stratified) consisted of 71 subjects from vendor. Such competition is possible only where new applications are concerned. A few low end x86 servers especially one-way servers are manufactured locally in India. especially for new workloads. z/OS commands less than 5 per cent market share. HP server with the Itanium processor is known as the ‘Superdome’ and is believed to have competencies that can match with IBMs mainframe/high end servers.cluster of ‘small servers’ may provide incipient competition to the high end servers. the older (legacy) applications and mainframe-only operating systems used on IBM mainframes are not compatible with RISC-based Unix servers or with x86-x64 servers. 11. high-end servers cannot provide competition for mainframes. Indian production is limited to the low end. Overall Windows server OS has the largest share in the entry level segment. Almost all of these servers are in the low range category of x86(Intel/AMD) servers. iv . Both IBM and HP competed for the SBI core banking contract which eventually was won by HP Superdome over the mainframe/P series configuration of IBM. very often the applications that run on Unix (such as Oracle database or SAP ERP) are compatible with the less expensive processors. do the PC and server manufacturing in India. Since use of application software varies across verticals. IBMs proprietary mainframe operating system. 8. Interestingly. chip makers. insurance and telecommunications. especially between verticals that support large enterprises such as banking. Therefore. of course. Although the Sun. Local Indian server vendors.

The data supports our conjecture which shows in case of HP. Linux supports only new workloads. The only exception to this general rule for high-end servers and mainframes is Linux. This shorter product refresh cycle indicates competition in the technology space. A successful adoption leads to positive feedback and others adopt the tested solution thus creating network effects. HP or Sun. One cannot choose to run Sun Solaris or HP-UX on IBM Power hardware. In the latter vertical IBM mainframes have a dominant position for legacy applications. as well as in other high-value verticals such as Financial Services. one of India leading integrated telecommunications operators. No single processor technology or OS dominates across all segments although clearly x86 and Windows OS is the preferred mix for the entry level server market. Reliance Communications still uses the Sun Microsystems platform for 70 per cent of its telecommunications applications. thus confirming the observation that legacy mainframe workloads are subject to monopoly lock-in on IBM mainframe. HP-UX and Solaris). Sun Microsystems had become very popular in the telecom vertical in the billing and customer care segments. If one buys a high-end RISC server from IBM. Pricing of a system includes cost of hardware. Our survey shows that all the vendors provide system software and management software with their hardware. larger share of installed OS goes to its own UNIX base OS. whereas in case of IBM larger share goes to its own AIX OS. while new workloads have typically gone to Unix or more recently Linux. Often hardware vendors bundle software with their hardware and charge accordingly. interoperability/openness and better service as important factors that differentiate their product from rivals. In fact. 13. In the high end market installed Linux is very low whereas share of proprietary OS provided by hardware vendors is comparatively higher. This finding however needs to be juxtaposed with the fact that in most cases vendors allow no other choice. 15. software and services. Likewise for IBM mainframes. Most vendors consider lower TCO. not legacy workloads. This implies that a priori the extent of ‘lock in’ for entry level users is likely to be less than for the other two segments. 14. In India vendors usually refresh their product cycle within 1-2 years. IBM is using its Unix servers (AIX on Power) to challenge Sun worldwide in the telecom market. Choice of processor and choice of OS varies across the segments. which can run on all of these types of hardware machines (though in the case of IBM mainframes. in most cases they only use IBM’s mainframe operating systems such as z/OS. Our survey reveals that users prefer a bundled purchase of hardware and OS in the enterprise level segment. As a result of this direct network effect. None of the vendors follow a vertical specific strategy although it could happen that a vendor may experience network effects in a specific domain. an increase in the number of users on the same network (platform) raises the consumption benefits for everyone on the network (platform). The aim of bundling software with the hardware is to reduce the amount of coordination v . For example.12. the only server OS choice is (usually) the variant of Unix sold by that vendor (respectively AIX. Competition in the technology space pushes firms to innovate and introduce new product lines. Moreover one cannot run Windows OS on IBM Power or Sun SPARC hardware (though one can on the Intel Itanium-based versions of HP Superdome). Suns dominance in the telecommunications vertical in India was broken by IBM which made a breakthrough in 2004 with a highly publicized US $ 750 million deal with Airtel.

In the case of Linux. which offers much lower profit margin opportunities). Unix and Windows Server switching costs reduce but do not eliminate the customer’s option to change vendors. This reflects the fact that once a customer has committed to a particular type of enterprise software (server operating system. Typically these discounts are offered on the upfront license price and not on the recurring maintenance fees. Only one user in our sample however was able to successfully negotiate a ‘once in a lifetime license fee’ for a specific software. 16.for a client across different vendors as well as saving on distribution costs. IBM mainframe operating systems and middleware are sold through a monthly license fee which varies with the size and usage of the mainframe hardware. China has emerged as a big market for the mainframe vi .). This is due essentially to the fact that: (a) the marginal cost of each additional copy of a software product is very close to zero. database. Indeed. Yet the market share for the IBM mainframe in the last 7 years worldwide has increased from 17% to 34%. etc. There is no one size approach in this regard in the industry and the actual price is the result of extended negotiation. The standard model for enterprise software vendors in most markets is to charge a single upfront fee known as a “perpetual license” and then to charge an additional recurring annual fee known as a “maintenance charge” (typically 15% to 22% per year of the perpetual license fee). On the other hand. The maintenance fee usually provides the customer with the right to receive all upgrades and new releases of the software. A key factor in this context – perhaps the most important from the vendors’ point of view – is that software enjoys significantly higher profit margins than hardware. and (b) the opportunities for proprietary differentiation are generally much greater with software than with hardware. Europe and Japan traditionally are big IBM mainframe users. US. which are now sold only through annual subscriptions. Oracle. mainframe has acquired 54 brand new customers across the world. In past one year. For example it could be per user. bundling often results in price discrimination and creates lock in for users. which is however fairly close in practice to a subscription model for most large customers. per core or per CPU or number of simultaneous users. It was thought that the mainframe would be rendered redundant by the faster computing power now available. others claimed to be charged according to a variety of methods. However. 17. SAP and IBM (for nonmainframe software) however adhere to the strict perpetual license + recurring maintenance fee model. And of course. Thus large system vendors such as IBM. middleware. all vendors offer negotiated discounts which depend on the bargaining power and long-term profit potential of the customers. vendors such as Dell who produce only server hardware and lack their own proprietary software offerings are far less profitable than the high-end system vendors (they can only resell software provided by third party firms. There are some exceptions to this rule. HP and Sun have a powerful incentive to bundle complex software such as server operating systems and middleware with their server hardware. Another exception is Sun’s Solaris (Unix) operating system and its Java middleware. the costs of switching to other kinds of software are quite high. Microsoft also has a somewhat different model. packaged application. Technological advancement in this rapidly progressing industry has meant that the power of the chip has exploded in the last decade. For example. of which 5 are from India. in the case of legacy workloads on IBM mainframes these switching costs are generally so high as to prohibit migration for all but the simplest (and generally least important) applications.

IBM’s relational database DB2. The mainframe market in India needs to be distinguished from the global market. ‘exit’ from a proprietary technology like the z/OS is indisputably more resource incentive (in terms of both time and money) than from a UNIX or Open source environment. That’s not a large base compared to mainframe customers worldwide and it reflects in the low revenue that IBM generates from mainframe sales in India. Thus even for HDFC the lock in factor is weak. It was based on a technical evaluation of the alternatives available. While all vendors try and create a ‘lock in’. But these new workloads represent a fairly small proportion of the total number of applications running on mainframes globally.10. Microsoft or Red Hat are less interested in lock-in than IBM. According to IDC there are 8.000 total IBM mainframe customers in the world. Unix and Windows Server are much more likely to share common standards and best practices (Java. but because the legacy applications and legacy operating systems (e. clean separation between data and programming logic. 18. programming methods and middleware were developed before many of these standards and best practices were fully evolved. HDFC claims acquisition of the Z mainframe for their credit card operations was independent of the prevailing system being used by the bank. IBM mainframes unquestionably have greater lock-in on average than Unix servers or x86 servers running Linux or Windows Server. Therefore for the five new mainframe customers in India. It is well established and widely known that while the server market ‘may be’ competitive ex ante. This greatly reduces switching costs compared to the mainframe. In India most mainframe sales are not upgradations or modifications. only two of the five users have opted for the z/OS. For the “new workloads” which can run on mainframe Linux or reduced cost mainframe “specialty engines” – and these would typically be applications that use Java (e. 19.g. The applications. Of these. The reason for low coverage in India. etc. where the operating systems. not just IBM. For “legacy” mainframe applications– mostly very large custom COBOL programs that mainframe customers have developed over years or decades – the lock-in effect is very strong which is why most sales globally are upgrades. Sun. lock-in is likely to be weak.lately. Mainframe is an enterprise class server which provides capability for extensive work load management and uninterrupted performance and security and thus suits extremely large workload requirements. only HDFC was using a legacy system AS-400.g. is that most companies do not have large data bases. IBM India disputed claims that the mainframe creates ‘lock in’. The vast majority of revenue generated by IBM from its mainframe business therefore represents upgrades to or replacements of existing mainframes.000. The total number of mainframe users in India including the newly added five is 25. the rest are on LINUX. But to be fair this is not because HP. The low installed base of mainframes in India suggests that new sales dominate in contrast to the global market. or certain packaged software such as SAP – the mainframe lock-in factor is much weaker. relational data models. it could be monopolized ex post resulting in high switching costs to the user. very few are new users. Most users were divided on the need for retaining flexibility (being able to switch to another platform to preempt lock in) and addressing their immediate requirements.). Interestingly. We did not find evidence in support of users choosing a particular platform based on the perceived ease of migration were vii . middleware and databases running on Linux. on IBM’s WebSphere application server). according to IBM. z/OS) running on mainframes are on average substantially older than those running on the more modern platforms. Of the five new customers in 2008.

the world's biggest network equipment maker has also recently entered the server market. 21. this is not an immediate worry for competition authorities. Oracle's planned purchase of Sun will combine software and hardware expertise. then IBM is a priori dominant and therefore needs to be tried and a remedy imposed under the law. We therefore need to further probe IBMs conduct and ask whether IBM has denied customers benefits of technological innovation and whether it charged above-market prices for IBM solutions. the Indian mainframe market is relatively young. including the mainframe in India. Windows applications too can run on servers from any of the leading server manufacturers. IBM’s reports show that about 2000 new mainframes were sold in 2008 but the vast majority of them were bought by customers who were already using mainframes. Our survey is optimistic about the extent of ex ante competition that exists in the Indian market. a multi vendor and multi platform approach. no longer believe that the relation between a high market share and market power is obvious. Ex ante competition in the market is no solace to a client who is locked in to a particular platform and post-purchase has very little option to migrate. there were only 54 new customers. there are somewhere between 8. legacy applications on the mainframe outnumber new acquisitions by a huge margin raising concerns about IBMs dominance in this segment. While there is some evidence for this in Europe and the US where IBM has a history of antitrust violations. On the other hand a mainframe application can only run on an IBM mainframe with IBM system software.5% of the sales. During the MRTP days this would have been sufficient to launch investigations against IBM because of its size. HP 33 and Sun 17 per cent of the market share respectively. In the US and Europe. Competition Authorities. In the US where the mainframe market is big and growing. viii . Our study provides little evidence of lock in because of some unique features of the Indian market namely a growing market. India has only 25 mainframe applications and only a handful of them run on the proprietary z/OS. the concerns in the US may become concerns in India as well as the base of mainframe users grows. and a late start in the computer age thus allowing leapfrogging many legacy applications. The z/OS is a proprietary system unlike UNIX or Linux. Linux or UNIX applications can easily be moved to virtually any hardware vendor’s platforms. On the other hand. Cisco. The SBI case study in the report provides evidence for this feature.they to become dissatisfied with system performance in the future. IBM holds 50. Instead choice was usually based on a vector of price and performance. Interestingly.000 total IBM mainframe customers in the world. For a market such as India. The benefits would be in the form of market expansion and perhaps lower prices especially in markets where the installed base of mainframe applications is high. well informed and knowledgeable clients. making it a formidable competitor to IBM.000 and 10. In India the high end market is dominated by the big three. 22. If we define this segment as the relevant market under competition law. But if we take a longer term view. licensing the z/OS would confer benefits all around. influenced by Chicago. albeit it is growing rapidly. Mainframe workloads can be moved to other high-end servers like Unix servers made by Hewlett-Packard and Sun. implying that a majority of the new mainframes were sold to existing users of mainframes. 20. mostly in the US and Europe. and IBM itself. representing 97.

23. in fact this would merely be a return to the decades-long practice of U. It would probably be much healthier for Indian IT and for the Indian economy if enterprises developed their complex business applications on server operating systems that are not tied to a particular brand of hardware. or Linux. This is nothing new. the study also finds that high-end server market as a whole in India is highly concentrated. Windows Server and to some extent Sun’s Solaris (which runs on Intel or AMD servers as well as on Sun’s proprietary SPARC servers. However. Today this is obviously the case with Linux (open source). Interoperability is a more complex issue but it is partially addressed in the two policy recommendations above. because mature mainframe emulation software products have existed for a number of years which allow z/OS to run on Intel server hardware. Such products include FLEX and TurboHercules. Because the Indian enterprise IT market is entering its high growth phase. 24. In fact. India would be much better off if CCI is able to successfully uncouple the lock-in strategy of the large server vendors before they come to completely control the market. to require that the sale of enterprise class server hardware shall not be tied to the sale of enterprise software (in particular. so we refrain from making it. but limits its use to software developers only and refuses to sell it to production customers. the leading Unix vendors such as IBM and HP do compel users of their Unix operating systems to purchase their brand of server hardware. the crucial server operating systems) is the first policy recommendation. Customers should be allowed to choose z/OS (or HP-UX. or Windows Server. The report finds that since there are few z/OS mainframes in India. If customers are allowed to purchase the server hardware of their choice for use with the server operating system of their choice (unbundled hardware and software) there may be nothing wrong with the fact that z/OS is not compatible with more modern operating systems. not just IBM) are required to license their software on RAND terms (Reasonable and NonDiscriminatory) These terms however would in no way restrict the right of vendors to earn fair revenue and profits from the sale of their proprietary software. India’s competition Authority (Competition Commission of India.e.S. etc. Requiring IBM to reveal its source code a la Microsoft is not going to be a pragmatic solution. CCI) has an excellent opportunity to avoid the pitfalls into which the more mature IT markets have fallen. IBM itself has also recently introduced such a mainframe emulator. and European regulators with respect to IBM as pointed out earlier in the report. This is our second policy recommendation. since HP’s SuperDome servers are based on Itanium processors. A competitive IT sector can confer substantial direct and indirect benefits to the Indian economy. Tying operating systems to a particular brand of hardware is the fundamental mechanism by which all the high-end enterprise system vendors – not just IBM – seek to lock in their customers and raise switching costs. But vendors should not be in a position to force customers to purchase only their brand of server hardware once the choice of server ix . Customer choice can be enhanced if server operating system vendors (all of them. It could theoretically be the case with HP-UX (HP version of Unix). It could even be the case for z/OS. 25.) if they wish. the consequences of IBM’s tying of z/OS to its own mainframe hardware (to the exclusion of would-be competitors seeking to offer IBM mainframe emulation on Intel servers) have been somewhat limited so far. Unbundling hardware and software i. on the basis of their own criteria of performance and functionality.

operating system has been made.g. India has an opportunity to leapfrog the more mature IT using economies by adopting a forward-looking regulatory stances that prevents IT system vendor lock-in from establishing itself as the norm in the Indian economy. To the extent that there is no inherent technical barrier to running z/OS on non-IBM hardware (e. customers should be allowed to make this choice. x . 26. on Intel or AMD servers using emulators such as TurboHercules). these are likely to be large indeed in the long term. Although we cannot quantify the economic benefits to India of such “pre-emptive” action. and rather significant even in the short term.

the structural dominance has translated into abuse of that dominance with IBM having to fight many antitrust cases in the US and other markets1. paras. A recent study of the European mainframe market has estimated that lack of competition in the market for mainframes will cost the European economy $48 billion over a 1 http://www. In the mainframe market or more generally the hardware. The mainframe therefore isn’t going to disappear any time soon. An efficient and competitive hardware and associated services sector has the potential to therefore generate substantial spillover benefits. For example. government etc) and between sizes of users. Introduction A mainframe is a high-performance computer used for large-scale and critical computing purposes which need greater availability and security than what a smaller-scale machine can offer. Supply side classification or segmentation will be complemented by demand side disaggregation to study whether there are any meaningful differences across verticals (i. The future of the mainframe. competition gains additional significance for two reasons. IBM has enjoyed a first mover advantage in mainframes with significant structural dominance in the market. while the market ‘may be’ competitive ex ante. The importance of competition in any market cannot be overemphasized. point 1. First.gov/atr/cases/f1200/1273. the result of over $1 billion of investment in R&D2. While the focus of the study is on the Indian market. and in February 2008 IBM unveiled its latest offering.The Issues of Competition in Mainframe and Associated Services in India I. because hardware and associated services are widely regarded as general purpose technologies that contribute significantly to the competitiveness of user industries. it could be monopolized ex post resulting in decline in welfare over a longer time horizon. the z10. and perhaps more crucially from the point of view of this study. although that distinction may be blurring as smaller computers become more powerful and mainframes become more multi-purpose.1.1 et seq 2 Gary Barnett. the source of contestability and the role for competition policy in this regard. This report will study the structural conditions in the mainframe and high end server market especially the extant barriers to entry. At times. Second. reasons for greater concentration in the high-end market will be analyzed and the impact of greater concentration on conduct will be studied. banking. Since their inception in the 1960s. the IBM mainframe celebrated its 40th birthday. Thus.htm and European Commission 14th Report on Competition Policy (1984). 94-95. EC 7/8-1984. Predictions of the death of the mainframe too have proved premature as IBM has reinvented itself to create sleeker and more powerful mainframes for the very large workloads it is capable of handling. especially those in the Asia pacific region and certain developed markets is done to provide useful insights. Historically. Bull. The mainframe and high-end markets will be compared with the medium range and entry level markets to identify differences in the structural conditions along with conduct in these separate markets. mainframes have been associated with centralized rather than distributed computing.e. a comparison with other countries. 2005 1 .justice. telecom. effects that are uncommon in the industrial economy such as network effects and switching costs are the norm in the information technology industry. In 2004. The question is whether the mainframe environment is likely to change over the next few years. operating system (OS) and application services market.

II. to make sure that products which currently exist only in laboratories can break into the market. is a task competition authorities will be well advised to pursue. Published data on price for different mainframe and O/S is not available 5 The Economist (Dec 18th 1997).e. (See Annex I for definitions). data constraints for the Indian market do not allow us to calculate these magnitudes with any degree of certainty.20 year period3. The goal of this study is however not to estimate the potential dead weight losses associated with the existence of firms with market power or ‘monopoly’ in the mainframe market. low-marginal-cost technologies commonly observed in these industries often lead to significant market power and second. We believe this is the first investigation of competition related issues in the server market in India. Even if products are getting cheaper and consumers are happy. while challenging. competition policy needs to address the issue of future competition i. this includes the operating system (O/S). innovation matters as much as price. In 1995 the Justice Department relied partly on similar reasoning to stop Microsoft buying Intuit. our approach relies on analysing structural dimensions for identified market segments using secondary data from Gartner6.cfm?story_id=E1_QVGSDN 6 Gartner is an information technology research and advisory firm. while reluctant to provide the exact price at which the deal was closed with the vendor. middleware and application software as depicted in Figure I.economist. revealed that discounts in the range of 50 to 90 per cent on the list price were not uncommon in this industry. Moreover. In fact a number of respondents we interviewed during the course of our survey. a personal-finance software firm5. Such price discrimination is important in high technology industries like the one we are investigating for at least two reasons: first the high-fixed cost. While this would be useful. Gartner’s data on vendor revenue is used to segment the market. since price will often exceed marginal cost. It becomes imperative to ensure that such costs are either eliminated or at least reduced by addressing the underlying reasons for the monopoly conduct. It is now widely accepted by competition authorities the world over including the Competition Commission of India (CCI) that structural dominance is however not a sufficient condition for exercise of market power. Developing a Framework for Analysis The market segment that we investigate in this study can be broadly characterized as enterprise-wide client server systems deployed in organizations to support operational computing. Given the difficulty in determining price precisely. Besides computer hardware. in technology industries. our secondary data assessment is accompanied by primary survey results of users across industry types and sizes to draw robust inferences of conduct across different segments. vendors will benefit from price discrimination. “Why Bill Gates should worry” Available at: http://www. 3 4 Jeff Gould ‘Benefits of Mainframe Competition for the European Economy’ January 15. Introducing effective competition in networked industries. Accordingly. 2 . The discount itself is a function of the short and long term worth of the client as determined by the vendor.4 For instance prices in this industry are typically not available in the public domain for India. 2009.com/ businessfinance/displaystory.

Annex II provides a summary of the vendors and the well corresponding market segment they serve in the Indian market. the intensive implications of these phenomena are relevant not only for corporate strategy but more orporate importantly for public policy. standards. who help clients evolve a workable IT solution by integrating the various ‘unbundled’ elements and often provide maintenance services as well. also known as independent software vendors (ISV). economies of scale. a segment in which it has had no presence until now.Components of Server market As shown in the figure. The economic phenomena that are important for high-technology industries and which are high highly relevant for the server market include personalization of products and prices. while a pure ‘vertical ‘box’ manufacturer such as Dell is a niche player devoted to making only computer hardware7. the client needs support or maintenance services to ensure smooth running of the computing infrastructure. An end to end solution provider such as IBM is a ‘vertical integrated’ entity serving the entire market. lock-in. but they are particular particularly important for technology-intensive industries. switching costs. Berkeley 2003 Varian 3 .computerworld. In addition to this. Most of these phenomena are present in conventional industries. and systems effects8. Finally there are systems integrators such as TCS (Tata Consultancy Services) and Satyam InfoTech. From the vantage point of this study. 7 Dell’s recent acquisition of system integrator Perot Systems will enable it to serve the high end server market high-end as well. Also present in this space are application software developers. bundling.com/s/article/9138343/FAQ_What_the_Dell_Perot_merger_means_for_the_IT_in dustry. OS developers such as Microsoft Windows and Red Hat Linux (or Novel Suse) that serve specific portions of the market. 8 See Economics of Information Technology Hal R. network effects. each logical layer is tightly integrated with the other to ‘produce’ services demanded by the client. Varian University of California. For more information on the merger please see http://www.

In this report we study the exercise. countervailing buying power and importantly the source of high market shares. causal relationship between the structure of an industry. 1982. (ii) over whether a correlation between concentration and profitability reflects differential efficiency or market power. their organization form and internal efficiency. 11 Demsetz. and (iv) over the implied assumption of symmetric effects. Inferring market power from structure was the logic however that led the Monopolies and Restrictive Trade Practices Commission (MRTPC) in India to focus on curbing 9 The interpretation and conceptual foundation of SCP studies has been questioned (i) over whether the relationship between structure and performance is expected in the long run or the short run. Firms differ in the products they sell. Market power is the ability to profitably raise price above marginal cost. one can envisage an alternative approach that makes the firm the centre piece of analysis. not the price takers of the perfectly competitive model. greater barriers to entry enable greater exercise of market power. (iii) over the assumption that structure causes performance. it is firm specific efficiency advantages that determine how large a firm grows and therefore industry concentration. extent of entry and exit barriers. however. Coca-Cola and PepsiCo.1985.10 It is the drive to be different that locomotes dynamic competition of the Schumpeterian sort. namely that the exercise of market power increases as concentration increases and two. control three-quarters of sales. This firm approach reverses the link between structure and conduct and performance. The traditional approach to assessing market power in the industrial organization literature is the Structure-Conduct-Performance paradigm (SCP). 10 Williamson. to take one example. “The Economic Institutions of Capitalism”. firm conduct. In contrast to this industry approach to conduct and performance. and market performance as measured by economic profits. We feel that the relation between structure and market power is far from being unambiguous. Typically the set of observable structural variables are measures of seller concentration and barriers to entry and the line of causality is envisaged to run from structure through conduct to performance or the exercise of market power. 4 . and effects of market power in the server market in India. We are however cautious about inferring dominance and its abuse from structural aspects since it ignores importance of competitors. The hypotheses of the S-C-P paradigm. “Barriers to Entry”. maintenance. Small numbers of competitors or the preference of consumers for a specific product bestows some degree of market power on firms. The implication is that concentration facilitates the exercise of market power9. America’s soft-drink industry. Thus more efficient companies with superior products grow to be larger than other firms. is noted for price competition although only two firms.11 It is difficult to apply either of these approaches in our present exercise since the focus is on a single industry (the server market). These features suggest that firms are price makers. We will begin with these hypotheses and test them using secondary data for the server market from India.During the course of the study it became apparent that this market possesses at least three important distinguishing characteristics: (i) products are highly differentiated and customized (ii) only a few relatively large suppliers exist in most segments and iii) the rate of innovation is very high. are of interest to us. rather than the traditional inter industry comparisons of performance and conduct of firms within those industries. The S-C-P approach assumes a stable. and competition is imperfect.

While it is tricky to draw a single conclusion from the mounds of case law. rather focusing on conduct. however. in a particular case. HHI varies between a lower limit of 0 and 1 (Monopoly) and the closer it is to 1. The inverse of HHI. Microsoft v Commission [2007] ECR.permit-raj” of 1970’s & 1980’s. prohibits and punishes only abuse of a dominant position. 13 United States v Microsoft Corp. By contrast. and the great mass of interpreting case law around it built over the years by regulatory intervention. the divergent decisions in the Microsoft cases in the US and Europe13.000. not entertain any new cases from now onwards.g. section 4 of the Indian Competition Act. This indicates that changes in HHI arise from changes in the absolute number of firms and the size distribution of firms. 2 the second largest etc. 1/HHI is the equivalent number of equal sized firms in the market that results in the same HHI. The “m” firm concentration ratio is the sum of the market shares of largest “m” firms 12 MRTPC Act 1969. the more concentrated the industry. Similarly. 1890. Civil Action No 98-1232 (DDC filed 18 May 1998) (Complaint) available at www. different competition authorities may reach different conclusions e. it is fair to state that over time anti-trust is less prone to punish big and successful companies in the US just because they are big and successful. It will. These are the Hirschman Herfindahl Index (HHI) and the concentration ratio.usdoj. the Competition Act 2002 steers away from structure as a determinant of market power. firms can become big through a host of legitimate business practices and many practices that appear on the surface to be anti-competitive can serve genuine procompetitive objectives. 2002.. The MRTP Commission will continue to handle all the old cases filed prior to September 1. Thus. The problem arises when there is abuse of dominance (AoD). MRTP Act is a grim reminder of the “licence-quota. then S1 ≥ S2 ≥ S3 ≥ Si ≥ SN . 2009 for a period of 2 years. Two measures of seller concentration are used. 5 . It does not condemn dominance per se.monopolies. The larger the variance of firm sizes-indicating a wider distribution of firm sizes around the mean-the larger the HHI. The Act had become redundant post July 1991 when the new economic policy was announced and Chapter III of the MRTP Act dealing with restrictions on M&A activities was made inoperative.gov/atr/cases/f1700/1763. Alternatively we order firms in descending order – firm 1 being the largest. HHI is also equal to 1 where σ2 is the variance of firm size.12 A monopoly was defined purely in terms of the market share which was then equated with dominance and its abuse. India’s modern competition law owes its intellectual debt to the US Sherman Act. Following the practice of the New Empirical Industrial Organization (NEIO) and the tradition established by competition law in the US.htm and Case T-201/04. AoD is however one of the challenging areas of competition law. If there are N equal sized firms.. then HHI= 1/N. these will be handled by CCI. If market shares are measured in percentages then the HHI is scaled by 10. we begin by analysing structural aspects of the server market in India insofar as structural dominance is necessary although not sufficient condition for establishing AoD by incumbents.

One is the idea that such industries are subject to “ever-increasing returns”. 6 . It is calculated as the sum of the percent market share of the top ‘n’ firms. products such as Windows.000) indicates an unconcentrated market A Herfindahl index between 0. if post merger HHI increases by 100 or more then scrutiny is done. with a very low concentration ratio. or C4. below 40% for the four-firm measurement. The next section defines segmentation of the server market from the supply and demand sides for the purpose of our analysis. Finally a comment about the nature of the market we are investigating. replacing software and so on. We explore both these ideas in our analysis of the Indian server market. Concentration ratios do not adjust.000 to 1.18 (or 1. above 60% for the four-firm measurement. present a high barrier against competitors’ entry. Two reasons why high-tech industries create special antitrust worries have attracted a lot of interest. C4 Limits Perfect competition. Any customer who wants to change to another product would need to spend a fortune retraining staff. If there is a winner in the race for technical standards. Section V analyses the market structure based on HHI and C4 estimates from secondary data. of the four largest firms in the industry. Section VII examines the issue of competition in the server market in India from a competition law perspective. buyer concentration may well make it difficult for sellers to exercise market power.1 (or 1. Monopoly. The second idea relates to “network externalities”. consists of the market share. it will not only to dominate the market but also create successful entry barriers. with its customer base of 100m people. For that we rely on evidence of firm conduct gathered from users and other secondary sources.18 (above 1. a review of the causes and consequences of regulatory intervention in the US market are invoked and lessons drawn for India. A finding of very high market shares however will not provide a robust assessment of AoD14. just as seller concentration is thought to be important because it raises prices. The concentration ratio of an industry is used as an indicator of the relative size of firms in relation to the industry as a whole.800) indicates high concentration.1 to 0. Monopolistic competition. FTC and Justice Department in US use HHI for evaluating Mergers. Section VI juxtaposes the results of Section V with those obtained from the extensive primary survey of Indian users. Since there is no history of cases against vendors in India. Four-firm concentration ratio. Section IV presents data for the Indian server market to establish its relative importance in GDP and also draws comparisons with Asia Pacific countries. Oligopoly. The rest of the report is organized as follows. It also offers a forecast of market size until 2012.800) indicates moderate concentration A Herfindahl index above 0. as the HHI does for variation in firm sizes. Purely by being so widely used. as a percentage.Commonly used measures are the four-firm (CR4) and the eight firm concentration ratio (CR8). Usually the assumption that buyers are unconcentrated is reasonable. with a near-100% four-firm measurement. However. Section VIII is on abuse of dominance by vendors in 14 HHI Limits A Herfindahl index below 0. We explore the nature and extent of countervailing buyer power in the Indian server market and the degree to which it provides a cushion against AoD by the vendor.

7 . the OS developed by Sun16. the end user. ii. The secondary data that we subject to analysis therefore covers computer hardware and the OS for the Apac (Asia Pacific. It is common in this industry that some applications are optimized for certain Operating Systems. Available at: http://www. and operating systems are useless without applications. Unix.g. There are at least five traditional ways of segmenting this market. goods whose value depends on their being used together. but our primary survey unambiguously points to the enormous heterogeneity in the server market. By number of processor sockets in the server. Windows Server. Not only do we feel that size is an unsatisfactory proxy for market power (for reasons stated above). Dell. Sun Microsystems has recently declared that it will develop Oracle in a manner that it works best with Solaris. HP. For example. Ideally therefore our relevant market should be a combination of computer hardware. or some intermediary. In addition. By server hardware brands. system integration could be done by the manufacturer. Fujitsu. there is no single universally acceptable ‘standard’ for segmenting the server market and it remains a thorny issue. Concentration in the server market could be compared with concentration in other markets and differences in conduct and performance could be attributed to differences in structure.oracle. Many of the users we interviewed for the study responded that choice of hardware and the corresponding OS was more often than not dictated by the choice of application. z/OS. By server selling price. has been optimized for z/OS i.Larry Ellison. By server operating system. iii. i. Annual data on application software is however not available since this could be shipped independent of the hardware.g. EPIC and CISC. price and nature of OS are too important to ignore. such as TCS or Satyam InfoTech making it difficult for a secondary source such as Gartner to capture application software data and map it to the computer hardware and OS data17. the IBM proprietary data base management system. IBM. By types of processors. In the server market hardware and operating systems do not by themselves add value for the user unless these are used together with application software. 18 Jeff Gould via email exchange.by 4 main types of processors: x86. 15 16 Hal Varian op cit Dramatically improving Sun’s system by tightly integrating Oracle software with Sun hardware. e. Having said that. we were frequently told that DB2. Linux. e.the presence of network effects and high switching costs. Coming to grips with the relevant market from the Supply and Demand Side It is common in high-technology industries to see products that are ineffective unless they are combined into a system with other products: hardware cannot be used without software. RISC.18 . Section IX provides concluding observations. Sun etc.com/features/suncustomers. OS and application software. including India) market. III. i5/OS.e. Differences due to technology.html 17 Gartner was unable to provide us with meaningful software data that we could map to the hardware and O/S. DVD players are useless without content. that is. If we were conducting a traditional S-C-P study. The economic impact of application software on the users is thus limited to the evidence we obtain from the primary survey. the bundle works best. then looking at the server market as a homogeneous entity would be appropriate. iv. v.15 These are all examples of complements.

none of the users considered interoperability to be a major factor limiting their choice.000 and above as high range. medium and large. or by number of processor sockets is not very useful for our purpose.e.Keeping the objective of the study in mind i. IDC. As we demonstrate later. if necessary. we decided to adopt the first two definitions. Moreover. IDCs high-end enterprise server range starts from US $500. choice of price bands may be disputed. then the 19 20 Worldwide and Regional Server 2008-2012 Forecast: March 2008. IT market research firms such as Gartner and IDC also classify the server market by average selling price and therefore this taxonomy makes it easier to compare our findings for India with the findings for the broader Asia market and the world market as a whole. Interoperability is a property referring to the ability of diverse systems and organizations to exchange information and to use effectively the information that has been exchanged. In our survey. The midrange and high-end markets are referred to collectively as the enterprise server market19. an examination of issues relating to competition in the server market. 8 .000-100. Table 1 presents the price intervals used in the study.e. On the demand side. Focusing on firms or brands is critical to any study of competition and average selling price enables us to slice the market into small.000 while IBM suggested we consider US $ 250. to create a different taxonomy. Table 1: Taxonomy of the market based on selling Price Prices (US $). there are fundamental differences between the small or volume market and the high end enterprise server market. or by type of processor used.000 as the cut off point for the high end configuration based on our finding that the Indian market is relatively more ‘price sensitive’. choice of specific computing infrastructure depends on the availability of applications. at current prices < 25000 >25. including mainframes Classifying the market on the basis of technology i. Whereas. but compatibility does not consider the efficiency of running a program on other systems.000 >100. We do invoke technology later in our analysis of competition. We eventually chose US$ 100. we assess the impact of rapidly improving chip technology and the emergence of cloud computing on the nature of competition in the server market. For example. In a market as broad and complex as the enterprise server market. We do not however use technology as an exogenous determinant of market classification. then it has no value. and vice versa. on nature of operating system. When network effects are present. multiple OS and application software can coexist in the market. If no one adopts a network good. For example. If there are enough adopters.e. For that to happen there needs to be interconnection or interoperability20.000 Entry level servers also known as volume server market Midrange enterprise server market High-end enterprise server market. there could be multiple equilibria i. Although both the terms are used interchangeably. compatibility refers to the ability of a program or hardware to function in the same way as other programs or hardware. we feel this classification is not limiting since the secondary data we work with allows us to subdivide these three categories into smaller bands or conversely merge price bands. so no one wants it.

IV. Sun Microsystems. we divide them on the basis of their employee size and type of application. A number of organizations in a wide array of sectors such as banking. they constitute natural target customers for leading vendors namely IBM. More detail on this is provided later. One could suppose that these rival vendors would make aggressive efforts to acquire new customers or to displace existing vendors. One user with multiple locations in India in the healthcare sector is in the process of migrating from standalone servers to a centralized facility to be located Hyderabad. especially in the public sector. A number of studies have shown that expansion of ICT sector not only contributes to GDP directly but also contributes to network effects (such as lower transaction cost. ITU 2006 Information Technology & Development: Beyond "Either/Or" by James Steinberg 9 . If there is a “critical mass” positive feedback kicks in and the product will be successful. ICTs have been recognized as a ‘General Purpose Technology’ (GPT) that transforms economic relations enhances productivity and creates new services and markets21. so more adopt it—making it even more valuable. Hence network effects give rise to positive feedback. specified ‘number of other users’ as part of their evaluation criteria for choice of OS and application software. As such. Indian Server Market: Prospects The new economics of industrial organization that has emerged over the past twenty-five years has recognized that particular industries have been dominated by networks either in terms of physical layout (identifying infrastructural links and nodes) or as virtual features through the provision of complementary products and services. Many of our respondents. a sector that demonstrates substantial network economies. Our focus group for the primary survey consisted of organizations using either the mainframe platform or enterprise level servers across many verticals including government. Like electricity.good becomes valuable. In our sample. the only product that ‘threatened’ to display such characteristics was ‘Oracle’. but of classifying their servers in terms of the applications they are running. manufacturing. efficiency and database management as well as improved customer interface. government. telecommunications and public utilities are using server systems and applications in India. transportation. High investment and usage of Information and Communications Technologies (ICT). The latter is not a way of classifying customers. But if the product never reaches a critical mass of adoption. retail distribution. There are scores of organizations in India that are looking to either upgrade their existing computing infrastructure or to install one for the first time. Microsoft and Oracle. health care. We examine this issue later. defense. Based on our primary survey we are of the opinion that this will only increase as the level of automation in a range of organizations increases. it becomes more or less entrenched. There is no doubt that pervasiveness of ICT coupled with its constant improvement spawns innovation and 21 22 World Telecom Development Report. Once a product achieves wide acceptance. it will inevitably result in failure. We do this not for the entire Indian market but for the companies covered by us in the primary survey. in mature markets has played an important role in their becoming information societies and reaping its benefits. Hewlett-Packard. rapid innovation) and hence increases the efficiency and productivity of user sectors and thus contributes to GDP indirectly22. These are naturally large organizations. The findings from the demand side analysis are reported in section VI. The reasons for the shift are improved security.

03) 46438 (6.2 (6. 2008.3 (10.547 (6.31) 7. According to NASSCOM this trend is likely to continue (see Chart1) 23 Breshahan T and Trajtenberg M (1995) General Purpose Technologies: Engines of Growth? Journal of Econometrics Jan 1995 Vol 65 No.437 (7.1 24 Nasscom-IDC study on Domestic IT Opportunity. Singapore at about 11 per cent sets the benchmark.740 (6.13) 10316 (11.93) 49.613 (10. Asia Pacific Years 2002 India 19.922 (10.67) 46.17) 11.258 (5. 35.400 (5.87) 54. The dramatic growth in Information Technology (IT) and Information Technology enabled Services (ITES) exports from India ‘overshadowed the latent opportunities unlocked and growth observed in the domestic market over the past few years’24.692 (5.676 (6.54) 26.376 (4.71) 75. Domestic demand for IT in India is witnessing a gradual transformation from being predominantly hardware driven towards a solutions oriented approach – resulting in a growing emphasis on services.09) Malaysia 6.95) 117632 (6.997 (4.99.56) 40.97) 8.02) 34.697 (7.06) 36164 (6.91) 6885 (7.25) 122.64) 8589 (8.947 (6.247 (8. Figures in Parenthesis are relative to GDP Recognizing the importance of ICT for development.34) (7.32) Philippines Singapore Australia 4.08) 8.77) Korea Rep.852 (6.751 (6. software and services) over the past few years25.43) 45. Servers and the applications that run on it are crucial to the smooth functioning of user industries.014 (10. many developing countries have increased such expenditure.273 (7.26) 9130 (7. although relative to GDP it is comparable with India. Table 2 above shows that China’s ICT expenditure is much higher than any other Asia Pacific country.83) China 62.17) 25. at times shifting investment away from other assets.92) 7863 (7.88) 58.443 (6. Table 2: Total ICT Expenditure US $ Million.509 (7.makes it easier to invent and produce new products or processes23. What’s more.19) 97658 (5.06) 40.62) 2003 2004 2005 2006 142313 10137.98) 4.17) 5502 (6.769 (6. The steep decline in ICT prices also encourages investment. 2006 25 Ibid 10 . revenue growth in the services segment alone has reported faster growth than that for the overall domestic IT market (including hardware.42) Source: World Development Indicators.63) 55304 (6.

Government. 2009-2014. due to infrastructure rollout in relation to the 3G initiatives are likely to be large buyers of IT services. and manufacturing verticals account for over 75 per cent of the revenues earned by service providers in the domestic IT services market. the slowing of the economy did affect India's server market (with an annualized decline of 29. however. Asia/Pacific. communications and media. post 2009 elections.27 Gartner expects the server market will follow closely on the heels of the economy. especially from telecom and banking verticals. 26 27 Ibid Dataquest Insight: Server Forecast Assumptions. the second half of 2009. We found support for this assertion during the primary survey as well. Gartner October 2009 28 Ibid 11 . and telecom. is expected to be relatively better. and the several large e-governance initiatives launched by the government under the National E-Governance Plan (NEGP) are expected to provide sustained growth in domestic demand for IT services over the next few years26. This is best demonstrated by the fact that most indigenous players in telecom and banking have significantly upgraded their levels of IT adoption to offer best-in-class services comparable to those offered by the global competition and these two sectors together account for approximately 35-40 percent of the domestic spend on IT services. financial services. Similar competitive pressures in other more recently deregulated service sectors such as airlines and insurance. According to Gartner.4% during the first half of 2009). At present. The Indian server market is predicted to grow at an average yearly rate of 10% annually until 201228.Chart 1: Server Market Growth Forecast Source: Nasscom-IDC 2006 Systems integration and network integration will be the prime drivers in the domestic market. and the uptake in the manufacturing and industrial sectors. Deregulation of key sectors has played a role in the progressively increasing IT adoption in the country.

h http://www.17% 10.org/external/country/IND/index.65% 5. In addition servers do need regular ervers maintenance services. Servers don’t last forever and typical replacement cycles are 5 years for all but the most basic of functions.Table 3: Revenue and Shipment growth for Asia Pacific 2008 Revenue Growth India Asia Pacific India Asia Pacific 12. 29 GDP is forecasted to grow at 5.12% 10. rather than with Singapore. thus reducing the gap with the countries immediately above it. According to experts.91% 7. the new range of server processors and their power efficiency is driving new demand. At the same time.29 Chart 2: Server market Growth as percentage of GDP Our primary survey revealed that one of the major impediments to adoption of servers is cost.18% -5. this is one of the few industries that has shown an exponential increase in performance with a constant decrease in prices.htm 12 .04% 7.71% 1. Many of the respondents stated that maintenance could be the leading cost item over the useful life of the server.org/external/country/IND/index.19% 7.73% 3. impediments Servers are costly to install and operate.58% 4.4% and 6.12% 4. See World Economic Outlook at http://www.81% -2.53% 7.64% 4. Given the growth forecast of Malaysia 10% for the server market.90% 4.90% 4.29% -1. As a result introduction and/or up gradation of server infrastructure is likely to be a carefully thought likely endeavour.78% 8.10% 5. Singapore’s server sector contributes the highest share of GDP.imf.28% 9.70% 6.21% 11.54% -3.57% 6.imf.67% 4. Australia and Malaysia respectively. which could be a significant cost item.64% 6.4% for 2009 and 2010 respectively. followed by China.04% 2009 2010 2011 2012 2013 2014 Shipment Growth Rate Source: Gartner Forecast As percentage of GDP (see Chart 2) the Indian server market still lags behind the comparator countries by a substantial amount. its contribution to GDP in India will increase over time.43% 6.

Operating system revenue is included but any non-OS software. V. distributed as a proportion across each reported sever-blade. begin their analysis of competition by asking a single question: what is the relevant market that is the subject of investigation. online transaction processing (OLTP) and running infrastructure applications. During the survey our respondents disclosed that information provided by Gartner can be considered as reliable and objective. 34 If there is only one mainframe supplier. This is not an exhaustive list. including all internal components such as system memory (RAM) and hard disk drives (HDD).33 In addition to defining product characteristics. even if the latter data were available it would not add value to the analysis.Our primary survey supports the inferences of IT research firms in that the Indian market represents huge growth potential with demand for computing infrastructure likely to emerge from enterprises and government. The data on servers available from secondary sources includes computer hardware and the OS bundled with it. The server market is one which is capable of performing a variety of functions including Web applications. Our examination of the Indian market did not reveal any region specific bias on the part of vendors. unless otherwise specified. 13 . economists. it is also crucial to define the geographic market especially from an assessment of competition point of view. Structural estimates of concentration (HHI and C4) are accordingly presented for the national level in this study. according to the formula for market concentration this just implies HHI = 1. it was an exercise fraught with error and risk31. But more importantly if mainframes and high-end servers occupy distinct albeit related markets then it implies a role for competition policy in the monopoly segment. This aspect is addressed in the policy section. but these server workloads cover the entire range of server functions of the respondents in our survey (See Annex III for description of each of these server functions). such as management tools and applications. Importantly this excludes the non OS software and although we attempted to collect information for the latter and map it back to the underlying system. high-performance computing (HPC). market and also work as a business consultant. e-mail/messaging. The quantum of data that needs to be addressed has grown by “leaps and bounds”30. if anything there may be vertical specific bias. our sample consisted of large organizations using enterprise level servers. Shipments data is number of servers sold. as stated earlier. processors (CPU) and network interface cards (NICs). We are unable to calculate structural estimates of market concentration for the ‘mainframe’ segment since there is no competitor to IBM in India34. front-end Webserving. Moreover data from Gartner on vendor revenue and shipments is available only on a national rather than regional level. revenue includes the cost of the blade chassis. Secondary data from Gartner is available for the period 2000-2009(1Q). In the case of blade servers. We therefore consider revenue and shipment data for all hardware internal to the server enclosure. In fact all respondents reported using most or all of these workloads since. streaming media. data warehousing. Gartner is a leading market research company which provides statistics on technology. In fact the only other manufacturer of 30 31 Interview with Intel Corroborated by Gartner and their analysts 32 See Gartner data 33 All secondary data used in this report is from Gartner. Structure of the Server Market in India and APAC No matter the issue at hand. This is a point we examine later in Section VI. as well as external storage is excluded32. A numbers of public and private organizations consult with Gartner prior to purchase.

35 Mainframe Executive.48 2. Many of our respondents were of the opinion that in terms of functionality HPs superdome and Sun’s SPARC system. 14 . among others.89 15. hardware platforms for new workloads (where mainframes and high-end servers can effectively compete). Every generation of the mainframe gets better. Joe Clabby.46 17. IBM had revenue of Rs 57. The data show that mainframe demand is extremely variable from one year to another. IBMs India-specific is estimated to be over $1 billion.42 20. 37 The z10 is Not Your Father’s Mainframe A modern assessment Terry Keene Integration Systems LLC 38 For an interesting and controversial debate around the demise of the mainframe see Terry Keene Op Cit and “The Story about the IBM Mainframe Makeover” www.44 Mainframe share (% IBM total revenue)_Apac 22. are different from legacy workloads.48 7. their sales aren’t big enough in India to permit a meaningful treatment on their own.80 trillion in fiscal 2007. not only for running financial market transactions but also to support millions of lines of programming code that are being developed for the rest of the mainframes circling the globe37. up from Rs 42.59 20.57 17. tasks that the mainframe performed earlier can now be efficiently executed by X86 based server platforms36. are capable substitutes for the mainframe39.33 20.com and others 39 See the SBI case study. more efficient.42 trillion in fiscal 2008 and Rs 33.32 20.com/articles/?p=107.66 3. Mainframe installations and skills are growing in India.48 11. faster.33 2. As a corollary.52 9.54 1. March 19. Indeed what defines a mainframe computer these days is vastly different from say its characteristics a decade ago because of technological progress.mainframes in the world today is Fujitsu. we therefore club it with the high end enterprise segment for the structural analysis38.mainframe-exec. However there could be an alternate and legitimate view that mainframes and high-end servers are in fact separate markets although as stated in the main text the data for India don’t permit them to be analyzed separately. for which mainframes do not have viable substitutes. Table 4: Share of Mainframe in IBMs Total Revenue for India.30 1.98 11. later for example. and carries more and more of the data load for business and industry all over the world35. In addition.33 7. 36 Interview with Intel September 2009.47 13.10 Mainframe share (% IBM total revenue)_China 20.83 trillion from domestic business in India for fiscal 2009. 2009 www.01 14.32 2. which does not yet market these computers in India (See Annex IV IBMs timeline in India). China and APAC Year Mainframe share (% IBM total revenue)_India 1. IBMs revenue from mainframe in India is minuscule compared to its overall revenue. Corroborated by IBM.21 2002 2003 2004 2005 2006 2007 2008 2009 Source: Gartner 2009 While it is premature to write the epitaph for the mainframe. India and China”. In the early 90’s there were no mainframes in India or China. As we show later.77 21. “Mainframe Sales Increasing in Southeast Asia.hp. on the other hand it is much higher in China and APAC (See Table 4).

192 0.173 0.318 0.379 0.279 High-End Segment 0.204 0.465 0.e.178 0.348 0.187 0.162 0.423 0.437 0.178 0.209 0.330 0.189 Mid range Segment 0. If the Herfindahl is low.347 0. the high end and medium range markets are persistently more concentrated than the low or volume market (see Table 6).269 Entry Level segment 0.336 0. The levels of concentration reduce slightly if we use shipment data instead of revenue.196 0.172 0.306 0.384 0.166 0.331 0. although the qualitative result does not change i.162 0.172 0.381 Estimates of C4 are not much different.262 0.320 0. a high Herfindahl.312 0.172 0.226 0.203 0.265 0.327 0. the high end market continues to be the most concentrated compared to mid and entry levels and interestingly.259 0. 15 .250 0.251 0. C4 for enterprise level servers (high and medium range) is close to 100.215 Mid range Segment 0.189 Entry Level segment 0. Table 5: HHI for Asia Pacific (APAC) by Vendor Revenue Years 2002 2003 2004 2005 2006 2007 2008 2009 Q1 Total market 0. Here we show the results for India and draw a comparison with China. on the other hand.313 0. The results for APAC are shown in Table 5.297 High-End Segment 0.341 0.223 0. results in a concentrated market in which price rises may be easier to sustain. there are many competitors and exercising market power should be hard.389 0.431 0.330 0.429 Table 6: HHI for Asia Pacific (APAC) by Shipment Years 2002 2003 2004 2005 2006 2007 2008 2009 Q1 Total market 0.275 0. The numbers show that the high end server market on average is more concentrated than the mid range market. 40 Estimates of HHI and C4 for all countries across the three defined segments are available in Annex V. We do this for both revenue and shipment data for the APAC region as a whole and for individual countries. The levels of concentration over time have not altered much either. Chart 3 shows that the enterprise level server market is more concentrated than the volume market and chart 4 shows that there are only 4 firms that exist in these two segments in India.426 0.402 0. which in turn is more concentrated than the entry level or volume market.196 0. This pattern is replicated in the revenue and shipment data for individual countries as well40. This implies that there are 4 firms that serve the entire market.165 0.259 0. which is determined by adding the squares of the market shares of all firms and C4 (market share of the top 4 firms in the segment.307 0.338 0.337 0.319 0.187 0.379 0.198 0.380 0.We measure concentration with the Herfindahl-Hirschman Index.

C4 is 100 for almost the entire period.35 0.50 0.05 0.20 0. We also test this argument based on our responses to the primary survey. For interesting both the high and medium range markets. Comparing the estimates for HHI and C4 for India reveal an interesting result.Chart 3: HHI for India by Vendor Revenue HHI for India by vendor revenue 0. while the volume or .30 0. while HHI fluctuates (see Chart 3) albeit within a narrow band.10 0. entry level market supports a larger number of firms and a correspondingly lower measure of concentration.40 0.00 2002 2003 2004 2005 2006 2007 2008 2009 Q1 HHI_entry level_India HHI_Midrange_India HHI_Highend_India HHI_Total market_India The nature of the result does not change when the analysis is done for shipment instead of revenue. The only significant difference between China and India is the lower concentration seen in China in the mid range segment in 2009Q1. the enterprise market is dominated by four firms. In China too. due to the entry of a new competitor in this segment.45 0. Chart 4: C4 for India by Vendor Revenue 100 80 60 40 20 0 2002 C4 for India by vendor revenue 2003 2004 2005 2006 2007 2008 C4_Entry level_India C4_Midrange_India C4_Highend_India C4_Total market_India 16 2009 Q1 . One can conjecture that the dominance of IBM in the high end market is being challenged (because its share in these segments has declined over time).25 0. The fact that the number (and identity) of firms in the enterprise segments does not change while the HHI fluctuates implies that change firms have traded market shares over the period.15 0.

it is possible to assert that application performance is as much a function of hardware as OS. These results show that variance on average is greater in the high end server market than the variance in the mid range server market. For example. reinforcing the earlier empirical finding that HHI is greater in the high end market than in the mid range and entry level markets. X86 processors offer flexibility to the customer in the choice of OS and applications. X86 is an open technology architecture that enables wider choices for the client. Only Sun Microsystems sold more servers with RISC processor than with x86. 44 For legacy workloads this will be true only if the applications and operating system running on the high-end server are compatible with the small servers in the cluster. In addition. Unix or Windows Server. 43 See Annex VII for distribution of processors by vendor across the three identified segments. by clubbing multiple small servers one can get performance akin to high end computing. We measured variance for the three segments for India. application software and other technical specificities41. 90 per cent of IBMs systems come bundled with x86 processors in the entry level market. the processor and the OS are often bundled with other hardware by vendors. especially for new workloads44. and the processor chip in IBM z10 mainframes) are designed for general purpose business computing. In other words. Depending on these needs. the RISC chips such as Sun SPARC. Although the Sun. We therefore limit our analysis to the choice of OS and choice of processor which will give us an indication of the ‘revealed preferences’ in the server market in India42. 42 In modern server operating systems such as Linux. the performance is closely tied to the OS. According to Intel. HP and IBM versions of Unix that run on their high-end servers are not compatible with cheaper x86-x64 servers that run Linux or Windows Server. High variance results in a higher HHI. The preferred processor for the entry level market is the x86 processor. At the entry level. For legacy applications such as those on the z/OS.Changes in HHI arise not only from changes in the number of firms and their market shares but also due to variance in the shares. one high end server can replace number of entry level servers. performance of x86 processors has improved manifold over the last 10 years providing benefits of “value and standardization”. database management software (industry experience says performance of database varies across the sectors). in fact Dell’s entire range of servers is sold with the x86 range. This indicates that a cluster of ‘small servers’ may provide incipient competition to the high end servers. Each processor family offers a particular set of price and performance points which overlap partially but not completely between families. however Sun Microsystems market share in India at 15% presently is quite low. it is nearly impossible to track all the information for each firm across each vertical. very often the applications that run on Unix (such as Oracle database or SAP ERP) are compatible with the less expensive processors. The two key building blocks of servers. 17 . This number progressively decreases to about 40 per cent and 0 for the mid and high end market segments respectively43. dominance of x86 processors can be attributed to the customer’s choice for open technology. the older (legacy) applications and mainframe-only operating systems used on IBM mainframes are not compatible with RISC-based Unix servers or with x86-x64 servers. which is greater than the variance in entry level market. Customers exercise their choice over a specific type of processor and operating system (performance of OS depends upon the number of processors in a multiprocessor environment and performance of application software largely depends on the choice of OS). the x86 processor dominates the entry level market across all vendors. a customer can choose a single vendor or opt for a multivendor arrangement. However. Across the three segments that we are studying. About 90 per cent of all IBM servers sold in India in 2009 included x86 processors. separately for revenue and shipments (Annex VI). according to the industry experts. Intel Itanium. Although it is important to analyse each preference type. 41 All of the processors considered here (Intel x86 & x64.

those based on Oracle or IBM relational database software). 47 Although there are separate OS for desktop and for server from Microsoft. While the Superdome’s hardware performance can match that of IBM mainframes for new workloads (e. 45 Windows share is 75 percent by shipment and 43 per cent by revenue. where these legacy mainframe applications are concerned.g. 2006-08 Share of CPU by Shipments. since Superdome cannot support these latter types of software. insurance and telecommunications. Since use of application software varies across verticals. e Therefore. But due to the paucity of data we limit our analysis of proce processor type to the 3 segments defined in our study.The same analysis on the basis of revenue rather than shipments shows that the share of x86 nalysis that x in overall revenue is about 55 per cent. Our survey of users unambiguously reveals that choice of OS is dictated by the application software. Market share for IA64 processor (Itanium) in HP machine has increased while that of the RISC processor in HP has declined since HP stopped ed selling servers based on its RISC chip in 2008. HP and Sun Microsystems with the latter a distant third. users of IBM mainframe operating systems and their associated legacy workloads are pretty much locked in. 46 We must again distinguish between new and legacy workloads. high-end servers cannot provide high end competition for mainframes.46 Both IBM and HP competed for the SBI core banking BI contract which eventually was won by HP Superdome over the mainframe/P series configuration of IBM (See Case Study in Annex VIII). while UNIX dominates the mid to high level segments47. Such competition is possible only when new applications are concerned. 2006 2006-08 IA64 Other RISC x86 1% 7% 16% 76% It is noteworthy that the high end market is effectively a triopoly between IBM. especially between verticals that support large enterprises such as banking. Overall Windows OS has the largest share in the entry level segment. here we are using Windows OS to mean OS for server only. 18 . This im implies the average price of the x86 processor is 86 lower than other processors45. HP server with Itanium processor is known as the ‘Superdome’ and is believed to have competencies that can match with IBMs mainframe/high end servers. Chart 5: Share of CPU by Shipment. We however seek answers to this question from the survey of users reported in Section VI. (See Chart 5 and 6) To understand this competition ideally one should analyse the demand for server in each segment of each vertical. choice of OS will also se vary across verticals.

In the mid range in case of HP. which can run on all of these types of end hardware machines (though in the case of IBM mainframes. Across segments the results are intuitive. HP-UX and Solaris). LINUX deployment has been increasing in the last 3 years and had 20 per cent market share in terms of shipments and 11 per cent market share in terms of revenue in 2008. in most cases they only use IBM’s mainframe operating systems such as z/OS. IBMs proprietary mainframe operating system. share of installed Linux OS is increasing over time. thus confirming the observation that legacy mainframe workloads are subject to monopoly lock on IBM mainframe hardware. The only exception to this general rule for high-end servers and mainframes is Linux. share of installed windows OS is declining48. 2006 2006-08 IA64 Other RISC 11% x86 2% 55% 32% Interestingly. larger share of installed OS goes to its own UNIX base OS. Our survey reveals that users prefer a bundled purchase of hardware and OS in the enterprise level segment49. 19 . If one choice buys a high-end RISC server from IBM. No single processor technology or operating system dominates across all 48 49 See Annex IX Users opt for a bundle of hardware and software from the vendor since it eliminates the possibility of conflict between the hardware and software and also makes it convenient to get service in the future. z/OS commands less than 5 per cent market share. your only server OS choice is (usually) end the variant of Unix sold by that vendor (respectively AIX. 2006-08 Share of CPU by Revenue. lock-in From the discussion above it is clear that choice of processor or choice of OS varies across the segments. Likewise for IBM mainframes. not legacy workloads. The data supports our conjecture which shows in case of d HP. whereas in case of IBM larger share goes to its own AIX OS. share of its own UNIX OS is declining along with the declining share of installed windows OS.Chart 6: Share of CPU by Revenue. HP or Sun. This finding however needs re to be juxtaposed with the fact that in most cases the vendors allow no other choice. Linux supports on new only workloads. In the high end market installed Linux is very low whereas share of proprietary OS provided by hardware vendors is comparatively higher. In case of IBM. Moreover one cannot run Windows OS on IBM Power or Sun SPARC hardware (though one can on the Intel ItaniumItanium based versions of HP Superdome). One cannot HP UX choose to run Sun Solaris or HP HP-UX on IBM Power hardware.

3 15.7 100 100 VI.3 100 100 2009Q1 100 20 10 70.segments although clearly x86 and Windows OS is the preferred mix for the entry level server market.0 100 25 0 50 25 100 100 2005 100 10.0 81. Sun Microsystems had become very popular in the telecom vertical in the billing and customer care segments. In India activities of server vendors are not region specific.0 0 100 50 0 33. We first report results for vendors followed by the user responses. For example. Primary Survey: Results Our primary data was collected by a sample survey conducted by ICRIER and TekPlus in India (see questionnaires in Annex X).1 0 100 21. Our sampling technique was purposively stratified and consisted of 71 subjects from vendors.4 0 52. Almost all servers sold by MNC vendors in India are imported from other manufacturing locations which implies that they are covered by the import duty regime. HP and to a lesser extent Sun Microsystems. Table 7: Percentage Share of Operating System by Vendor in the High End Market by Shipment Vendor Hewlett-Packard Linux Others UNIX Windows IBM AIX Linux UNIX Z/OS/OS390 Sun Microsystems UNIX 2002 100 0 25. Indian production is limited to the low end. of course.6 0 57.0 60. for example HCL has its manufacturing capabilities in Chennai and Pondicherry.3 100 100 2003 100 0 42. namely IBM. Local Indian server vendors.0 0 100 53.2 100 100 2007 100 0 18.9 17. A break up of the subjects is provided in Annex XI.0 75.7 23.0 75.4 100 29. service providers and a representative cohort of medium and large enterprises from different verticals. Reliance Communications still uses the Sun Microsystems platform for 70 per cent of its 50 See Annex XII for duty structure and the corresponding SIC codes 20 .3 53. Some large MNC vendors leverage their desktop/notebook capabilities to produce the low-end volume servers. The goal of this primary survey was to collect qualitative and quantitative information from vendors as well as users across verticals.7 14.1 30.0 0 100 48.1 0 63.8 5. Our choice of users was restricted to medium and large enterprises since the aim of this study is to examine the nature and extent of competition in the high end market which is largely served by the ‘Big Three’ in India.3 16.9 13. A few low end x86 servers especially one-way servers are manufactured locally in India. We pursue this argument in Section VII.3 100 100 2008 100 0 25.0 13.8 10.6 100 100 2006 100 0 19. This implies that a priori the extent of ‘lock in’ for entry level users is likely to be less than for the other two segments. chip makers.2 15.8 100 100 2004 100 13.2 81.6 21.0 0 100 28. do the PC and server manufacturing in India. In fact.1 30.9 57.8 0 100 50 0 26.1 14.50 None of the vendors follow a vertical specific strategy although it could happen that a vendor may experience network effects in a specific domain. Almost all of these servers are in the low range category of x86(Intel/AMD) servers.5 6.

an increase in the number of users on the same network (platform) raises the consumption benefits for everyone on the network (platform). while new workloads have typically gone to Unix or more recently Linux. Most vendors consider lower TCO.51 As a result of this direct network effect. The standard model for enterprise software vendors in most markets is to charge a single upfront fee known as a “perpetual license” and then to charge an additional recurring annual fee known as a “maintenance charge” (typically 15% to 22% per year of the perpetual license fee). which are now sold only through annual subscriptions. HP and Sun have a powerful incentive to bundle complex software such as server operating systems and middleware with their server hardware. Another exception is Sun’s Solaris (Unix) operating system and its Java middleware. For example. The aim of bundling software with the hardware is to reduce the amount of coordination for a client across different vendors as well as saving on distribution costs. In India vendors usually refresh their product cycle within 1-2 years. which is however fairly close 51 As is indeed the case worldwide – Sun has always had significant share in the global telecom segment. and (b) the opportunities for proprietary differentiation are generally much greater with software than with hardware. as well as in other high-value verticals such as Financial Services. There are some exceptions to this rule. one of India leading integrated telecommunications operators (See Bharti Case Study. which offers much lower profit margin opportunities). Six out of eight vendors also provide application software and four provide security software. Our survey shows that all the vendors provide system software and management software with their hardware. IBM mainframe operating systems and middleware are sold through a monthly license fee which varies with the size and usage of the mainframe hardware. Indeed. Suns dominance in the telecommunications vertical in India was broken by IBM which made a breakthrough in 2004 with a highly publicized US $ 750 million deal with Airtel.telecommunications applications. The maintenance fee usually provides the customer with the right to receive all upgrades and new releases of the software. Competition in the technology space pushes firms to innovate and introduce new product lines. software and services. In fact IBM is using its Unix servers (AIX on Power) to challenge Sun worldwide in the telecom market. This indicates that main objective of the product refresh cycle is to provide better technology which can meet future demand at cheaper price. A key factor in this context – perhaps the most important from the vendors’ point of view – is that software enjoys significantly higher profit margins than hardware. Pricing of a system includes cost of hardware. Thus large system vendors such as IBM. In the latter vertical IBM mainframes have a dominant position for legacy applications. This is due essentially to the fact that: (a) the marginal cost of each additional copy of a software product is very close to zero. This shorter product refresh cycle indicates competition in the technology space. price competitiveness and market vision.Annex XIII). Microsoft also has a somewhat different model. vendors such as Dell who produce only server hardware and lack their own proprietary software offerings are far less profitable than the high-end system vendors (they can only resell software provided by third party firms. bundling often results in price discrimination and creates lock in for users. A successful adoption leads to positive feedback and others adopt the tested solution thus creating network effects. Often hardware vendors bundle software with their hardware and charge accordingly. Our survey reveals that factors considered important in product refresh cycle by vendors are technology demand. On the other hand. The network effect here is thus a truly global and not only a local phenomenon 21 . interoperability/openness and better service as important factors that differentiate their product from rivals.

especially in the presence of countervailing buying power. In the mid range. the costs of switching to other kinds of software are quite high. vendors usually offer a buy back option. Our survey of vendors reveals that 5 of 8 vendors offer unlimited licensing which means the client has to pay only once to use the software. Typically these discounts are offered on the upfront license price and not on the recurring maintenance fees.). 22 . In this case hardware vendors’ deal with the client for the software installed in the system but for any services required the client needs to contact software vendor. According to vendors.in practice to a subscription model for most large customers. Unix and Windows Server switching costs reduce but do not eliminate the customer’s option to change vendors.52 In the high end segment competition is less not only because of fewer players but also because of greater customization of products. Hardware vendors may have an agreement with the ISV (Independent Software Vendors) and charge for the software installed in their hardware from the client. Introduction of new product line makes the existing product outdated. To induce clients to purchase new products. The greatest number of vendors perceive high competition in the entry level market. packaged application. Table 8: Extent of Competition in the three segments Market segment High End Mid Range Entry level 52 Number of vendors (scale of 1-10. middleware. SAP and IBM (for nonmainframe software) however adhere to the strict perpetual license + recurring maintenance fee model. profit potential) of the customers. as emphasized right through this study smaller numbers in the high end market does not necessarily indicate lower intensity of competition. in the case of legacy workloads on IBM mainframes these switching costs are generally so high as to prohibit migration for all but the simplest (and generally least important) applications. However. Oracle. Only one user in our sample however was able to successfully negotiate a ‘once in a lifetime license fee’ for a specific software. database. 10 –low) 1 2 3 7 2 0 0 0 3 2 5 1 4 0 0 0 5 2 0 0 6 1 0 0 At entry level 7 vendors ranked level of competition as 1(highest) whereas at mid range 3 ranked it as very high (1) while 5 others ranked it as 3 on a 10 point scale. the extent of competition in the higher end is the least while it is the highest in the volume segment. competition is expectedly more than the high level segment but less than that perceived in the entry level. etc. This reflects the fact that once a customer has committed to a particular type of enterprise software (server operating system. In India 5 of 8 hardware vendors offer leasing option and 7 out of 8 offer buy back options while offering new products. all of the vendors offer negotiated discounts which on the bargaining power and long-term desirability (i. 1-high. We examine this point later. others claimed to be charged according to a variety of methods. There is no one size approach in this regard in the industry and the actual price is the result of extended negotiation.e. However. per core or per CPU or number of simultaneous users. And of course. For the moment we report that respondents perceive competition is lower in the mid to high range compared to the entry level segment. The survey also revealed that vendors usually lease the older generation products to the lower end market but do not sell refurbished servers in India. Table 8 below reports vendor perceptions. In the case of Linux. For example it could be per user.

since Oracle is likely to have a much larger share of the database market on non-mainframe servers (especially Unix). According to IBM these volumes are not possible in Linux.53 In past one year. only two of the five users have opted for the z/OS. The total number of mainframe users in India including the newly added five is 25. mainframe has acquired 54 brand new customers across the world. In addition. Europe and Japan traditionally are big IBM mainframe users. of which 5 are from India54.An interesting characteristic emerging from our analysis of vendor data relates to the perception that competition is lower in software services as compared to maintenance and hardware services. In practice. China has emerged as a big market for the mainframe lately55. For both these types of products the sales practices and extent of lock-in potential are very similar for IBM and Oracle. It’s however important to note that IBM has a much broader product line than Oracle (at least prior to the anticipated completion of Oracle’s acquisition of Sun). It was thought that the mainframe would be rendered redundant by the faster computing power now available. Another reason for this could be that the entry level server market has to a large extent been commoditized and may have influenced the responses.ibm. 10 –low) 1 2 3 4 5 7 0 0 0 0 1 0 2 1 3 3 0 3 0 1 6 0 0 0 Technological advancement in this rapidly progressing industry has meant that the power of the chip has exploded in the last decade.com 55 Bank of China is a good example of a recent acquisition for IBM. IBM and Oracle can reasonably be compared for two specific software products: relational databases and Java-based middleware. Only 2 vendors mentioned that government policies are not conducive to reducing price for end users and one of them mentioned it is not creating a level playing field. Table 9: Intensity of competition in the submarkets Market segment Hardware Software Maintenance (including services) Number of vendors (scale of 1-10. 53 54 Gartner data From www. it is not surprising that customers perceive its software to possess greater lock-in effects than similar software from IBM. Many user respondents corroborated this feature claiming that ‘Oracle’ tends to create a greater ‘lock in’ than does IBM. z/OS is used for the core banking solution delivering 9. 1-high. Interestingly.box in real-time based on more than 380 million accounts with 3 billion transaction histories. is that most companies do not have large data bases. Mainframe is an enterprise class server which provides capability for extensive work load management and uninterrupted performance and security and thus suits extremely large workload requirements. the rest are on LINUX. 23 . z/OS is most cost-effective in such situations. according to IBM. The reason for low coverage in India.445 business transactions per second (tps) on one. That’s not a large base compared to mainframe customers worldwide and it reflects in the low revenue that IBM generates from mainframe sales in India. Yet the market share for the IBM mainframe in the last 7 years worldwide has increased from 17% to 34%. However. 6 of 8 server vendors mentioned that they do not collaborate with their competitors. US.

For training students on mainframe platfom and automating server production with virtualized servers on System z. The organization has consolidated from 96 cores of Intel Xeon servers to eight IFLs running at about 3040% peak utilization. Reliability. if any The workloads were unable to be completed in time on other hardware. HDFC Bank System z10 EC 216 MIPS. Scalability and Security of applications and data. This is now a real time application. Oracle license reduction to almost one-sixth. z10 BC for DR Z-OS Batch window problem in other environments could not be solved.expresscomputeronline. Consolidation/ Virtualization. Source: http://www.Table 10: Mainframe Clients added in 2008 in India Organization Bajaj Allianz Details of System z used for deployment System z10 BC with 1 IFL and 6 Capacity Backup IFLs.com/20090413/market01. 1 z9 BC with 7 IFL processors OS Linux Workload deployment Bajaj Allianz in India has deployed its data warehousing applications alongside DR on System z. Coimbatore has becomes one of the world's first educational organizations to offer a Masters in Technology program with a Mainframe Technology specialization at the same time it has installed a new System z z10 BC running z/OS and Linux to support university workloads while offering direct student access to the university's technology engine. Satyam Anna University z9 BC with 26 MIPS.shtml 24 . simplicity in management. Performance of batch applications has improved. EL-COT System z9 EC . Performance was a major factor Power consumption is contained even if the workload grows to 15 times the current volume of transactions and data.4IFLs and 64 GB Memory in each machine Linux One System z is deployed in the data center and another one is used for mirroring all hosted applications at the DR site. Credit Card Application Wants to be the hub for all colleges in Coimbatore for MF training and also wants to start virtualization and consolidation around zLinux. and 3 IFL processors Z-OS Linux Anna University. Continuous availability. Benefits TCO benefits due to SW license reduction.

Microsoft or Red Hat are less interested in lock-in than IBM. clean separation between data and programming logic. This greatly reduces switching costs compared to the mainframe. etc. lock-in is likely to be weak.g. Therefore for the five new mainframe customers in India. Sun. The low installed base of mainframes in India suggests that new sales dominate in contrast to the global market. one must distinguish between users choosing a platform for a new application (workload).g. IBM’s relational database DB2. According to IDC there are 8. The mainframe market in India needs to be distinguished from the global market. For the “new workloads” which can run on mainframe Linux or reduced cost mainframe “specialty engines” – and these would typically be applications that use Java (e. ICICI opted for a non mainframe solution citing huge price differential.10. But to be fair this is not because HP. It was based on a technical evaluation of the alternatives available. 57 AS/400 is a proprietary IBM mini-computer. The vast majority of revenue generated by IBM from its mainframe business therefore represents upgrades to or replacements of existing mainframes. Based with almost similar requirements. IBM mainframes unquestionably have greater lock-in on average than Unix servers or x86 servers running Linux or Windows Server. The applications. not a mainframe although it’s not unreasonable to lump it together with mainframes for the purpose of this study. shortage of technical skills and limited application software compatible with the mainframe ecosystem58. We did not find evidence in support of users choosing a particular platform based on the perceived ease of migration were they to become dissatisfied with system performance in the future.60 56 Interview with system integrator. Of the five. very few are new users.IBM naturally disputed claims that the mainframe creates ‘lock in’. where the operating systems. Instead choice was usually based on a vector of price and performance. Of these. and users who own large legacy applications that were created many years ago using custom mainframe software tools. z/OS) running on mainframes are on average much. new sales are more likely to be new acquisitions rather than upgradations. much older than those running on the more modern platforms. For “legacy” mainframe applications – mostly very large custom COBOL programs that mainframe customers have developed over years or decades – the lock-in effect is very strong which is why most sales globally are upgrades. programming methods and middleware were developed before many of these standards and best practices were fully evolved.000.57 HDFC claims acquisition of the Z mainframe for their credit card operations was independent of the prevailing system being used by the bank. ‘Exit’ from a proprietary technology like the z/OS is indisputably more resource incentive (in terms of both time and money) than from a UNIX or Open source environment59. 60 Again.000 total IBM mainframe customers in the world. But these new workloads represent a fairly small proportion of the total number of applications running on mainframes globally. middleware and databases running on Linux. on IBM’s WebSphere application server). One of the myths that got broken (in India) by new sales was that most mainframe sales are upgradations or modifications56. but because the legacy applications and legacy operating systems (e. relational data models. Most users were divided on the need for retaining flexibility (being able to switch to another platform to preempt lock in) and addressing their immediate requirements.). 58 Interview with ICICI Bank 59 We examine the impact of Open Source later in the study. However since the installed base in India is low. 25 . All vendors try and create a ‘lock in’. Unix and Windows Server are much more likely to share common standards and best practices (Java. another major private sector bank. not just IBM and it arises due to technology was how one user put it. only HDFC was using a legacy system AS-400. or certain packaged software such as SAP – the mainframe lock-in factor is much weaker.

CRM etc. Telecommunication. Utilities. This indicates that there is no brand specific bias in purchase. each deployment being decided on a ‘case by case’ basis. In our sample of firms we came across only one user who switched from IBM to HP following dissatisfaction with performance. 26 . 62 This fact can be inferred from Tables 11 through 14 read along with the profile of the companies in our sample in Annex XI.) Core Applications (Banking. Of the 41 users we covered in diverse verticals such as Healthcare. ‘lock in’ to a specific technology or vendor can be mitigated by a multi vendor and multi OS approach. Insurance and Government almost all the users put their server infrastructure to a wide variety of workloads while making use of multiple vendors and multiple technologies62. and therefore the primary form of mainframe lock-in is likely to occur only in a small number of cases. Details of the migration are therefore not available. Process Manufacturing. For example 33 of 41 firms use multiple vendors. Table 11: Vendors providing servers to the organization Brand IBM HP Sun Microsystems Dell HCL Wipro Fujitsu Other Number of users 29 30 23 5 5 2 0 3 Table 12: Main uses of Server infrastructure Types of Workload Web hosting Mail server File & Print Server Security Applications IT & Network Management Cross-industry Applications (ERP. Banking. Retail. it could be monopolized ex post resulting in high switching costs to the user. there are very few customers in India. some were of the view that this helped to create competition among vendors.61 Apart from migration which is a ‘last resort’. Although none of our users seemed to be explicitly following this particular route. telecom etc. 61 Numerous attempts to contact Vishal Mega Mart were unsuccessful. Transport.It is well established and widely known that while the server market ‘may be’ competitive ex ante.) Messaging & Collaboration Others Number of Users 25 39 36 19 24 33 35 30 1 In the latter class.

63 The reason that there are no standard benchmark results comparing the performance of IBM mainframes with Unix RISC machines or Linux/Windows Server x86 machines is that IBM mainframe sales contracts expressly forbid the publication of any such comparative data. 12 or roughly 30 per cent refresh when the need arises (see Table 15). A result of high price (even if we were to accept the high price hypothesis) does not inevitably translate into anti competitive conduct or AoD. We did find support for this claim during our interviews. The modal refresh cycle in our sample is 5 years or more. IBM disputes that their prices in India are high. What we were able to ascertain though is that significant discounts ranging from 50-99% on the list price are offered. which owns an IBM mainframe for testing purposes. rather than short term.00 100 Table 14: Operating Systems running on the servers Types of OS Open Source Only Proprietary Only Open Source and Proprietary Number of users 0 15 26 Lock in or a high exit barrier for users also manifests itself in high price. The extent of discount varies across clients. we came away with a strong sense that Indian buyers were knowledgeable about the market and were technically well informed. Interaction with different vendors did not help either. The reason for the perception that price is high is due mainly to the focus in India on total cost of acquisition (TCA) rather than on total cost of ownership (TCO). Many reports have suggested that IBM-mainframe prices are relatively high compared to other similar systems delivering the same functionality. On the contrary.Table 13: Server architecture organization is predominantly using Architecture X86 RISC EPIC/Itanium Other Total Percentage of users 97.07 0. Our attempt to test the higher price hypothesis with figures however was thwarted due to paucity of data on actual “deal” price. Our survey. Further such price information is not publicly available. While this may be true in the case of small enterprises. it certainly is not true for the enterprises we surveyed. According to IBM.93 17. cross-platform benchmarks citing that each customer situation is different and has to be studied independently. did not suggest that buyers focus only on TCA. Even Microsoft. IBM provided no data based on standard. As one user said “they will charge what the market can bear”. is not allowed to publish any benchmark data for that machine. conducted as it was among large enterprises. It is a fact the Indian server market like others is highly customized and prices are negotiated between the clients and vendors. 27 . We take this to mean that server infrastructure investment is made for the long. and frequently use them for competitive marketing purposes against HP. TCO for them is low because of scale. Sun and Dell. Note that IBM’s ban on publication of hardware benchmark results concerns only their mainframes. They publish many benchmarks for their Power-based Unix servers and their x86-x64 servers.56 82. although it was at times countered by statements like “there is no competitive benchmark to the IBM mainframe”63.

More than 100 million accounts are processed daily in SBI. but are perhaps only a minority of the much smaller and younger mainframe installed base in India. Two. Lastly. Interestingly. This is certainly not true for our sample. we know elasticity is a function inter alia of the number of substitute’s available and countervailing power buyers are able to exercise in the market. 65 See Case study (Annex VIII). According to one of our respondents. The assumption of zero cost for OSS for servers. They claim that OSS fails when number of transaction/instruction crosses a certain limit hence question its reliability and scalability. Three attributes stand out in this context. For critical applications majority of our survey respondents (21) use 51-75 percentage of their server infrastructure to run mission critical application whereas 10 clients use 76-100 % of their server infrastructure to run mission critical applications (Table 16). private sector firms and educational institutions. The analysis is based on 20 case studies of Indian organizations drawn from government departments. It is this latter group of customers who are the victims of IBM mainframe market power. This is because it needs to be put through a grueling test cycle before implementation and requires maintenance support. knowledgeable Indian clients are able to oppose exercise of market power by any vendor including IBM. They constitute the vast majority of the worldwide mainframe installed base.Table 15: Frequency of upgrading / replacement of the existing server infrastructure Years Less than 3 years 3 – 5 years More than 5 years As and when need arises 3-5 years and more than 5 years Number of users 0 7 18 12 4 Although we do not have an exact estimate of the elasticity of demand in this market. It is widely believed that open source software can be considered as an alternative to the proprietary software to reduce cost and increase transparency. none of our respondents rely exclusively on open source. The conclusion of the study is mainly based on guesstimates. Industry experts mention that customers use OSS on parallel on trial basis but rely on proprietary platforms to run mission critical application. Other criticisms of OSS articulated by our respondents ranged from it being suitable for research universities rather than for an organization running mission critical applications to its fluid nature being disadvantageous for large firms. India missed the computer age and is a late starter in this context. vendors. Open Source Software (OSS) has an important and lasting development role to play especially in emerging economies. and coupled with the fact that it is a high growth economy. including IBM have little incentive and ability to exercise market power. 28 . if any. One. Coopers and Lybrand once famously remarked IBM works best in an environment where the hosts know little64. A recent study conducted by IIM Bangalore analyzed the economic impact of FOSS (free and open source software) in India. OSS comes bundled with a cost and “overall TCO of running Linux is no different”. HPs acquisition of SBIs core banking contract on the superdome platform demonstrates that banks can use a mainframe alternative for their core processing65. Therefore. at the 64 Negotiating skill of knowledgeable Indian clients’ vis-à-vis IBM may not be relevant to the small number of Indian users who have older (legacy) mainframe applications. As a result the number of legacy mainframe applications is few. departs from reality.

There are several such mainframe emulators available. because high profits will attract new competitors. What this implies is that the mainframe (based on IBM’s proprietary z/Architecture and using IBM’s proprietary z10 microprocessors) actually participates in two separate computer markets which are defined by their operating systems: the legacy mainframe market (z/OS). it is possible to switch from mainframes to high-end servers and vice-versa. Table 16: Percentage of server infrastructure used to run mission-critical data Percentage 1 – 25 26 – 50 51 – 75 76 – 100 Number of users 0 10 21 10 VII: Analysis of Competition in the High End Server Market in India: Is it Adequate? Competition policy and economic regulation are based on the premise that the “public interest” or “social good” is best served when markets work efficiently. The Chicago School economists left a legacy of free market principles that defined competition policy for many years at least in the US. The server market. Identified with the Chicago School. Sun or IBM itself is much more challenging. Hercules)67. On the other hand. either by acquiring them (PSI). because they are tied to an operating system (z/OS) that cannot run on these servers. or threatening them with patent litigation (T3. Second. See the policy section for more detail in this regard. 66 67 Robert Bork: The Anti Trust Paradox. However. for legacy workloads. they said. First. 29 . and one segment of the high-end server market (Linux). governments should stop worrying about size and ask only whether a firm can exert market power. Hence. If markets work efficiently then economic or regulatory intervention can do more harm than good66. the shift from mainframe to high end servers of HP. the effect will usually be temporary. Let’s look at structure to start with in the high end market. Also see Annex XIV for antitrust litigation against IBM in this matter. The questions we are investigating in this study are whether IBM has exercised market power in the server market in India by charging high prices and has it denied customers the benefit of innovation. The conditions for the markets to work efficiently however are so stringent that in practice they are almost never satisfied even at the best of times. It makes sense to look at the high end broader market and not just mainframes since customers can switch to other high end servers at least for new workloads. with high entry (and exit) barriers and tremendous network effects can scarcely be described as the best of conditions for markets to work efficiently.same time majority ( 26 out of 41) of them run mission critical applications on proprietary OS while using open source to run less critical applications. Note that it is in fact technically possible for z/OS to run on servers with Intel/AMD x86 or x64 chips using software emulators. all quite mature and with excellent performance. Legacy mainframe workloads cannot switch to high-end servers. as this study has been arguing. In the case of new workloads such as Java or Linux-based applications. even if a firm gains market power. markets will erode most monopolies more quickly and effectively than will governments. IBM refuses to license z/OS for use on these emulators and has acted to suppress them.

30 . as a result of I.79 44. Sun Microsystems and Cisco confirm this.M.00 45. A glance at the advertising and R&D expenditures of IBM.00 0.00 0.50 0. One might hypothesize that IBM’s comparative underinvestment in R&D is the behaviour of a monopolist. During the MRTP days this would have been sufficient to launch investigations against IBM because of its size.85 1. While serving as an entry barrier.B.72 10.00 28. A summary of these is provided in Annex XIV. although the reality must certainly be more complex than this.91 100 2004 0. Sun spends almost 15 per cent on R&D.39 44. R&D cuts both ways.00 16.39 43. In fact IBM has a history of antitrust investigations against it in the US.07 2.54 0. followed by IBM which spends a little more than 6 per cent. Microsoft and Intel. given that competition for the market is almost non-existent because of the high entry barriers69. since we know that the legacy mainframe business (monopolistic) is only one part of IBM’s business. no longer believe that the relation between a high market share (or a high HHI) and market power is obvious.00 2006 0. IBM had almost 50 percent market share. including India.79 48. We therefore need to further probe IBMs conduct and ask whether IBM has denied customers benefits of technological innovation and whether it charged above-market prices for IBM solutions. “customers have been denied the benefits of technological innovation and must instead pay above-market prices for I.73 33.00 33. Tichy.73 0. R&D helps bring the benefits of innovation to the market place.89 35. with HP holding around 33 percent share and Sun Microsystems 17 percent (see Table 17).00 0.00 2008 0.” 69 Some experts claimed that Cloud computing can become a competitive threat for incumbents.00 12.47 28. it is unlikely to create much competitive pressure a the high end.25 2009 Q1 0. Mainframe workloads can be moved to other high-end servers like Unix servers 68 In a paper commissioned by Microsoft examining the alternative mainframe technologies. Whether and in what form the innovation reaches the customer is in part determined by the extent of competition in the market.68 These are the very questions that IBM is being investigated in the US at present. two large companies with high market share and pervasive network effects also have a history of antitrust cases and violations in the US.At the high end.00 44. Only a company like Cisco can overcome these barriers.45 7.47 0.00 0.M.’s actions.81 0.90 0.71 36. Walter F.48 0. Our survey is optimistic about the extent of ex ante competition that exists in the Indian market.65 17.97 100 100 36.58 46. mainframe solutions and premium wages for a dwindling mainframe workforce. Table 17: Market Shares in the High-end Segment Vendors \ Years Fujitsu/ Fujitsu Siemens Hewlett-Packard IBM SGI Stratus Computer Sun Microsystems Total 2002 0. and that other parts (non-mainframe software and services) are in highly competitive (albeit oligopolistic) markets.57 0. a professor of computer science at the University of Karlsruhe in Germany.B.87 100 100 There can be no doubt that there are huge entry barriers in the high end server market worldwide.91 48.90 100 29.47 18.08 40.00 35. influenced by Chicago. Competition Authorities.00 2005 0. concluded that. At the high end however due to security features absent or lacking.56 100 2003 0.40 100 2007 0.25 1.93 0. including the mainframe in India.

mainly those based on Linux or Java. Oracle's planned purchase of Sun will combine software and hardware expertise. What about price? Our study is less clear cut on this issue.21 398455 260. The SBI case study provides evidence for this feature. and IBM itself70. In the concluding section we report a price comparison between India and China for the mainframe based on secondary (Gartner) data with interesting results.26 386558 255.64 98786 6153 6. the world's biggest network equipment maker has also recently entered the server market.47 25 0. While revenue sharing deals between IBM and Airtel take initial price out of the equation71.75 2008 13880 1834 13. and a late start in the computer age thus allowing leapfrogging many legacy applications.18 34200 405. Our study provides little evidence of lock in because of some unique features of the Indian market namely a growing market.08 39540 5325 13. Cisco.made by Hewlett-Packard and Sun. cannot be easily moved to Unix servers (or to Linux or Windows Server machines) 71 Idea and Vodafone are expected to sign similar contracts with IBM 31 .19 61535 56.12 1259 1.98 66129 59. prices in general were not available for the Indian market.17 7401 21. In US a switch takes time and is costly particularly in the case of custom applications written specifically for IBM mainframe. Table 18: R&D and Advertising for the Big Three Sun Microsystems Net Revenues ($ million) R&D expenditure ($ million) R&D expenditure as a % of revenue Advertising expenditure ($ million) Advertising expenditure as a % of revenue Number of employees Revenue per employee ($ thousands) IBM Revenues ($ million) R&D expenditure ($ million) R&D expenditure as a % of revenue Advertising expenditure ($ million) Advertising expenditure as a % of revenue Number of employees Revenue per employee ($ thousands) CISCO Net sales ($ million) R&D expenditure ($ million) R&D expenditure as a % of Net sales Sales and marketing expenditure ($ million) Sales and marketing expenditure as a % of Net sales Number of employees Revenue per employee ($ thousands) Source: Company websites 2007 13873 2008 14.23 1242 1. well informed and knowledgeable clients. for reasons already discussed. Prices for the mainframe in India 70 Again. Legacy mainframe workloads.21 32 0. this is only true for new workloads. a per chance multi vendor multi platform approach. making it a formidable competitor to IBM.79 Ex ante competition in the market is no solace to a client who finds himself locked in to a particular platform and after purchase has very little option to migrate.71 103630 6337 6.55 34922 4598 13.47 8690 21.23 34900 397.

IBM made the decision to discontinue all licensing of mainframe software for use on non-IBM systems (Annex XV). The ones who did saw nothing improper in it since it is IBMs IP. 75 See Annex XIV 32 . On the other hand a mainframe application can only run on an IBM mainframe with IBM system software. IBMs strategy simply revolved around protecting its mainframe monopoly. When we asked IBM the reason. How they did this is described ahead. 72 73 Jeff Gould op cit Benefits of Mainframe Competition for the European Economy’ January 15. Windows applications too can run on servers from any of the leading server manufacturers. Few years ago. In the US where the mainframe market is big and growing. The proposed remedies are cast within the framework of competition law in India. Abuse of Dominance in the Presence of Network Effects and High Switching Costs Old habits die hard. Linux or UNIX applications can easily be moved to virtually any hardware vendor’s platforms. we were told that the Z/OS can only run on IBM mainframes and they were not aware that the OS used to be licensed a few years ago! Users mostly refused to comment on this aspect. it points to a strategy in which IBM is using price discrimination to penetrate new markets while continuing to “lock in” older markets. The z/OS is a proprietary system unlike UNIX or Linux. In the US and the EU this resulted in frequent antitrust actions against IBM75. There can be little doubt that IBM has in the past engaged in a concerted series of actions designed to erect barriers to mobility and entry against competing providers in the global market and hence protect its market power from a variety of threats74. At the beginning of this section we observed that ‘social good’ is best served when markets work efficiently. Rather than a taking a static approach which involves looking at this issue with reference to only what obtains in this market in India at present. with its widely accepted principle of consumers’ welfare being the only meaningful gauge. competition policy. This evidence is however only suggestive and more data from these markets would be necessary to conclusively prove this hypothesis. this is not an immediate worry for competition authorities. 74 A barrier to entry is a cost borne by an entrant and not borne by an incumbent that tends to discourage competition. In case they do not. The question however arises how this can be achieved and more importantly whether we are overplaying the dangers of monopolisation of the mainframe market by IBM.73 The benefits would be in the form of market expansion and perhaps lower prices especially in markets where the installed base of mainframe applications is high. we take a slightly longer term perspective to identify any potential competition concerns in this growing market and suggest remedies for improving overall welfare. This is not a question just for the Indian market but one that concerns the entire server platform world over.are comparatively lower than for China and combined with evidence from Europe where IBM seemingly acts monopolistically72. But if we take a longer term view. VIII. For a market such as India. should support a judgment in favour of opening up the mainframe market. And finally the more difficult question of proprietary technology. Moreover none of the mainframe or high end IBM buyers in our sample actually felt ‘over charged’ by IBM. the concerns in the US may become concerns in India as well as the base of mainframe users grows. licensing the z/OS would confer benefits all around. 2009 Analysts estimate that a quarter of IBM’s US$100 billion in revenue comes from the mainframe.

Microsoft owns the operating system standard and predictably dominates the market. entrenchment of a particular standard may be temporary. April 2000 http://www. a reference was made to the applications barrier to entry78. market forces would win out and therefore there is no need for an antitrust law to police such practices. to Apple’s Macintosh”79. or penalized that customer in some other way. in the celebrated antitrust case against Microsoft in the US. the appeals court held that Microsoft went beyond the boundary of normal commercial practice. In the judgment against Microsoft. “have an overwhelming incentive to write for Microsoft’s Windows as opposed. Once a product receives wide acceptance. and since Microsoft dominates the desktop operating systems market. while for Linux it is 1%.nabe. charged a higher price. In the desk top market.htm 79 http://www. the alternative platform has been unable to make much headway in the desktop operating systems market.pdf 80 Robert Litan. This meant that if any customer—whether an Internet service provider or a computer manufacturer—dealt with Netscape. Microsoft Windows won the battle for desktop OS and became the dominant standard. Microsoft overwhelmingly dominates and has been able to successfully frustrate competition. But any actions that entrench the monopoly are liable to be punished. Business Economics. 78 http://www. Legal and Economic Aspects of the Microsoft Case. the fact that Microsoft withheld or threatened to withhold its operating system Windows or what are called “applications program interfaces” (APIs) from other competitors was considered to be a manifestation of exclusionary conduct76.While the law in the US (and India) allows creation of a monopoly. But along with additional activities such as exclusionary conduct—Microsoft was successfully able to further entrench the applications barrier to entry and in fact raise it. a principal competitor. applications writers. There is a broad-reaching presumption that Microsoft’s hold on the market stifled innovation over the long run80. the verdict stated. Since applications writers have to write their programs to mesh with operating systems.nabe.com/publib/be/000245. A firm could conceivably charge a price that the market would bear without actually violating the law.justice.77 The larger question in this case was the barrier to entry issue. Due to rapid innovation. for example. Microsoft withheld Windows. 77 Market share for Windows is 95%. it proscribes certain practices that have no commercial justification in an effort to further entrench the monopoly. Linux remains a distant third and Mac OS at second with 5% market share. That in itself is not illegal. For example. therefore becomes “for the field” rather than “in the field”. It is claimed that 76 APIs are important in the operating systems market because they are the hooks that allow applications programmes to mesh with operating systems software. it becomes more or less entrenched. One of the arguments against breaking up Microsoft—it was one of the proposed remedieswas that for certain applications. firms with market power need to be cautious about exercising it. These sorts of activities are precisely those prohibited for a firm with market power. Technologically dynamic markets like the one we are investigating are characterised by network effects which means that one product or standard may tend towards dominance since the utility a user derives from consumption increases with the number of other users consuming the good or service. Competition. in such cases. Breaking up Microsoft would have fractured the standard. Linux. Specifically. Some economists however contend that in the long run. Under competition law.gov/atr/cases/f3800/msjudgex.com/publib/be/000245. consumers like the fact that there is a single standard.pdf 33 .

that is. Romer. a maker of computer chips. Although it may be desirable to wait and see if a monopoly proves only temporary before acting.85 The Microsoft and Intel cases are crucial in establishing that antitrust has a role to play in the “new economy”. it is highly likely that competition will result in “fragile monopolies” being created. AMD. Another difficulty is to predict how the sector will develop in the event of antitrust action against a dominant firm. pp. In many instances involving new technologies. Accessed from http://www.25 billion in cash. it may be difficult to uncover transgression since a degree of temporary monopoly may be part and parcel of innovation.82 On the other hand. No. Because technology such as software requires huge fixed investment up-front. allow its rival to outsource the production of its chips (which are based on Intel patents) and abide by a set of “business-practices provisions”.innovation has the capacity to alter the field altogether81. out of the market84. 5. Vol. independent of an industry’s structure83. Intel’s rebates kicked in if customers gave the firm between 80% and 100% of their business that were often conditional on the exclusion of its rival.com/businessfinance/displaystory. Long ago. by Paul M. inefficient monopolies may not survive for long. "Part 2: The Problem of Development: A Conference on the Institute for the Study of Free Enterprise Systems. 84 The Unkindest Cuts. the evolving market in India and suggest ways for tackling this thorny issue based on secondary and primary 81 Constance Bagley: Note on Application of Antitrust laws to the New Economy: An Analysis of United States v. 1990).cfm?story_ id=14885938&fsrc=rss 34 . In this model of the “new economy”. Against this background we evaluate IBMs mainframe ‘monopoly’." (Oct. it has also been observed that the notion that antitrust enforcement is not needed in the new economy rests on the erroneous assumption that technological change is exogenous.economist. in contrast to monopolies in the old economy. Striking the right balance therefore will be a key challenge for the CCI (Competition Commission of India) if it seeks to investigate the mainframe market in India for current or potential abuses of competition. it is not right to dispense with antitrust enforcement in the new economy. Journal of Political Economy. Joseph Schumpeter argued that the pursuit of monopoly power would drive innovation and become the central driving thrust of the ‘new economy’. Oct 5th 2000 From The Economist print edition 83 Endogenous Technological Change.06 billion ($1. The EU feels that Intel used rebates to price AMD out of the market. In return AMD will drop all litigation against Intel and withdraw all of its regulatory complaints worldwide. with single companies dominating segments for a time.44 billion) for illegally using its muscle to price AMD. The conduct of Microsoft in seeking to stifle innovation and of Intel in pricing AMD out of the market are particularly blatant examples of AoD. Yet the Microsoft case suggested that a sufficiently powerful incumbent could act as a barrier to innovation. Following an investigation into its alleged anti competitive practices. until they are toppled by rivals. 71-102. Microsoft Corporation. but involves trivial marginal costs. The recent antitrust finding against Intel. a rival chipmaker. Aug 20th 2009. 98. In fact antitrust laws have a builtin premise that the wait can be too long if companies have market power combined with a willingness to exercise it. is a case in point. The Economist print edition 85 Intel in an out of court settlement will pay AMD $1. the European Union (EU) in May 2009 fined Intel €1. Harvard Business Review September 2001 82 The new enforcers. Antitrust action can thus reduce the exercise of market power and potentially improve the welfare impact.

Lack of options also inflates service and maintenance costs in these markets. 35 . an application running on the mainframe is difficult to port to another system or standard. its actions had come under the scrutiny of the Department of Justice (DoJ). Indeed IBM has been prosecuted for anticompetitive tying and exclusionary conduct in both the US and the EU. For example. one gigabyte of memory for an IBM mainframe costs US$6.e. The question was whether IBM was using anti-competitive and illegal tactics to monopolise the business computing market. “network effects” make it hard for rivals to enter an industry already dominated by an incumbent with an established network of users. It is useful to compare the sources of market power for Microsoft and IBM respectively. IBMs source of market power is largely linked to its proprietary operating system standard which is not compatible with the products other system vendors. UNIX or Windows server costs less than US$20088. The proprietary nature of the OS and the corresponding lack of interoperability with other vendor systems makes the ‘switching costs’ very high. IBM was also investigated by the Directorate General for Competition at the European 86 IBM was forced by the Department of Justice to license mainframe operating system software to other hardware vendors. including by high price and expensive service contracts.87 In these markets mainframe products cost substantially more than comparable products from non-mainframe server providers.org/ research-resources/understanding-ibms-mainframe-monopoly. On the other hand. IBM is not new to antitrust. allowing the sale of its mainframe software only with its mainframe hardware – a practice also followed by Microsoft when it ‘bundled’ its browser with the operating system90. As IBM became more and more dominant in the computing industry in the second half of the twentieth century. which ties the customer to expensive service from IBM89. The legal proceedings that followed led to a consent decree with the United States DoJ which imposed conduct remedies on IBM in the mainframe market. due to the proprietary nature of the z/OS the possibility of customers moving off the mainframe to other platforms is not easy. While network effects were paramount in the case of Microsoft. IBM stopped all such licensing in 2007. the charge against IBM was one of tying i. Table 20 compares the key elements that have made these two pieces of intellectual property. IBMs closed and incompatible system creates an ‘application barrier to exit’ for the client i. The attendant ‘lock in’ of the customer can be exploited in any number of ways. As stated above.86 In addition. since IBM requires all mainframe customers to have IBM maintenance agreements for their mainframe software.000.html 88 Ibid 89 During our Survey we were informed that dwindling mainframe skill sets in US and EU are also responsible for high costs.information we gathered for this propose. 87 See for Instance “Understanding IBMs Mainframe Monopoly” Available at http://openmainframe. Network effects allowed Microsoft to create the “applications barrier to entry” that was a big element in the case against it. while the same amount of memory for a Linux. 90 Bundling and tying are related. Meanwhile.e. Microsoft used its muscle in the OS market to effectively destroy Netscape in the browser market by bundling its own browser internet explorer with Windows. in the mainframe market IBM allows sale of its mainframe software (the z/OS) only with its mainframe hardware – a concept known as tying. See fn 71 also. On the other hand. A client on the mainframe platform would therefore face higher ‘switching’ costs compared to being on a platform which is compatible with others. a practice frowned upon by most competition authorities. Microsoft’s almost total dominance in market for desk top operating systems resulted in more and more applications being written for the ubiquitous Windows thus creating what is now called the ‘application barriers to entry’. Microsoft’s Windows and IBMs z OS dominant in their respective market segments.

91 Table 19: Sources of Market Power for Microsoft and IBM Compared Microsoft Windows PC Desktop System 90% Very Strong High Application barrier to entry. Through its actions. ii) offer an unbundled version of its mainframe computer processing units without main memory and iii) allowing the sale of its system software for use on other manufacturer’s mainframe computers. It continued licensing its older 31 bit software until 2007.Commission which led to an Undertaking by IBM in 1984. After the migration from 31 bit to 64 bit mainframe hardware and software in 2001. prices have remained high and customers have fewer choices than if there were more vendors 91 92 See for Instance “Understanding IBMs Mainframe Monopoly” op cit. This refers to IBM’s refusal to continue licensing its older 31 bit mainframe operating systems for use on the FLEX mainframe emulator formerly resold by T3. IBM never subsequently agreed to license its 64 bit operating systems for use on non-IBM hardware. the consent decree in the US and the Undertaking in Europe. Department of Justice. With no competition in the mainframe platform market. ban on exclusive dealing and license z/OS to other hardware & software vendors Source: Authors’ compilation * If we consider the entire server market. IBM has prevented alternative mainframe solutions from being viable and as a result it now controls 100% of the mainframe market. which coincided with the end of the DoJ Consent Decree. IBM was initially induced to license its mainframe operating systems for use on other vendors’ machines by pressure from the U.92 As a result IBM plug compatibles have disappeared from the market giving IBM a virtual monopoly in the mainframe segment and therefore the ability to charge more for its mainframe hardware. the relevant share becomes 100% The mainframe has now again become increasingly proprietary. Thus IBM agreed to i) supply interface information for its mainframe computers in a reasonable and non-discriminatory manner (RAND). 36 . where application writers would choose to write for windows because of 90% market share Unbundling Operating System from browser. were aimed at reducing the extent of tying in selling of mainframe computers and creating a more open ecosystem around the proprietary standard. Both actions.S. IBM mainframes constitute roughly 5% of the total server market (by revenue). Since we have argued earlier that IBM z/OS mainframes in fact belong to a unique market. A large and competitive market for IBM-compatible mainframes from vendors other than IBM existed for 25 years (from approximately 1975 to 2000). huge barrier to exit for clients Operating System Hardware Market Share Network Effects Switching Cost Nature of Barrier Remedy Provide Interface Information (API). Two years ago. ban on exclusive dealing IBM z/OS Mainframes 100%* Weak Enormous Once on mainframe cannot port easily. IBM made the decision to discontinue all licensing of mainframe software for use on non-IBM systems.

How this will be done is a moot point and we do not get into the nature of the remedy here. The vast majority of these purchases represented upgrades to or replacements of existing mainframes corroborating IBMs monopolisation of the legacy mainframe market96. It is possible however to find a parallel of such a circumstance in the telecom market. it is plausible that by far the greatest portion of IBM’s approximately $5 billion in 2008 mainframe hardware sales (and virtually all of its $9 billion in mainframe software sales) came from existing customers. our interaction with experts in the field leads us to the inescapable conclusion that mainframes are servers in a class of their own. these skills are becoming increasingly scarce and expensive. 37 . implying that a majority of the new mainframes were sold to existing users of mainframes.000 and 10. on IBM’s WebSphere application server).5% of the sales.94 For legacy applications that run only on the mainframe and where the switching costs are high IBM charges high price to perhaps ‘cross subsidize’ markets that are more competitive. for legacy applications mainframe customers have little choice. legacy applications on the mainframe outnumber new acquisitions by a huge margin raising concerns about IBMs dominance in this segment. availability serviceability. If we define this segment as the relevant market under competition law. In fact IBM has developed an elaborate price discrimination scheme which sets different prices for the very same mainframe hardware depending on whether the intended application is locked-in legacy (unable to migrate) or a so-called “new workload” (that could run anywhere)95. Interestingly. Notwithstanding the monopolization of the mainframe market by IBM.000 total IBM mainframe customers in the world. IBM’s relational database DB2. Mainframe professionals who write applications or administer mainframe systems have unique skills that take years to develop and are different from skills required for any other platform. The mainframe offers unparalleled reliability. In the US and Europe. For so-called “new workloads” which can run on mainframe Linux or reduced cost mainframe “specialty engines” – and these would typically be applications that use Java (e. as indeed you begin to do in the second paragraph following the table.g. It is crucial to distinguish unambiguously these two cases.creating and selling alternative solutions. In Europe and the United States. analyst from Gartner Ibid. An implication of this pricing scheme is that while high end servers from HP and Sun Microsystems can provide competition to ‘new’ mainframe workloads. performance and security93. Therefore. then IBM is a priori dominant and therefore needs to be tried and a remedy imposed under the law. IBM’s reports show that about 2000 new mainframes were sold in 2008 but the vast majority of them were bought by customers who were already using mainframes. Mobile call termination is often defined as the relevant market by many telecom regulators around the world and regulated since termination is 93 94 PJ Koul. or certain packaged software such as SAP – the mainframe lock-in factor is much weaker. According to data available from IDC there are somewhere between 8. mostly in the US and Europe. The applications use unique languages. Deputy General Manger of IT Department of Air India and Massimo Pessanini. 95 Jeff Gould 96 For so-called “legacy” mainframe applications – mostly very large custom COBOL programs that mainframe customers have developed over years or decades – the lock-in effect is very strong. For new workloads IBM sells its mainframe hardware at discounts exceeding 90% of the price it sets for legacy workloads. there were only 54 new customers. systems and interfaces that simply don’t exist together on other platforms. What is interesting however is whether defining the relevant market as “legacy mainframe market” will pass legal muster. representing 97.

98 The main reason for such a high estimate is the large installed base of IBM legacy applications in Europe. mobile termination is regulated as a monopoly.considered a monopoly. Interestingly. The enterprise IT market as a result is still young. See ITU publication on this. This study provides some insights into how real these costs could be and how these can be mitigated. Lack of inter-operability (a proprietary standard) and a restrictive licensing policy together generate a closed system. If this is common knowledge among clients they will be cognizant of the risks of going down this inflexible path and therefore take decisions that allow for low cost course corrections in the future. Indeed this will be true of any market that has a sizeable installed base of mainframe legacy applications. IBM itself is offering an Integrated Facility for Linux (IFL) on its mainframe hardware. the lock in and corresponding dead weight losses will be lower. Certain exogenous developments in the server space such as cloud computing and government procurement in India could change the nature of competition within the server industry. IX. Conclusion: Future Market Development. op cit 99 Ibid 38 . A large number of applications running on the proprietary z/OS cannot migrate to other operating systems because of technical complexities99. As the Indian market grows and server deployments increase across the range of government.97 The similarity between the mainframe and the telecom market is in the nature of ex ante and ex post competition. where the installed base of mainframe legacy applications is small. It is useful to recall the reasons for the high costs associated with mainframe legacy applications. This would for example involve choosing an open standard such as Linux or Unix to begin with. public and private enterprises. While the mobile market is largely competitive ex ante. Combined with IBM’s restrictive policy on z/OS licensing this means that clients are effectively tied to IBMs z system hardware to run z/OS software and applications.8 10. albeit growing very rapidly. Although z/OS’s cumulative market share in India is small. was a relative late starter in stitching together IT systems for its companies. despite its recognised prowess in producing skilled information technology professionals. which has outsold the z/OS sales in India in 2008. On the other hand in a market. IDC and NASSCOM cited in the report are optimistic about future deployment of IT infrastructure in public and private enterprises and in government. we discuss these aspects and the corresponding implications for conduct of future competition policy keeping in sharp focus IBMs conduct in Europe and the US in the mainframe segment. call termination is regarded as a monopoly because of the lack of alternatives in terminating a call on a unique mobile number. In the next and concluding section of this report. 98 Benefits of Mainframe Competition for the European Economy Jeff Gould January 15 2009 v6. Competition Policy and role of CCI It is indeed fortuitous in some sense that India. This is quite similar to the situation observed in the mainframe market as described above. In fact IBM has shipped more open standard mainframe systems in India than with the proprietary z/OS overall resulting in a low cumulative market share for the z/OS. both clients and the regulator in India need to be aware of the potential risks of lock in to a proprietary standard. such as India. severance from which involves prohibitive costs. industry experts are convinced that its advanced features and exceptional performance is unquestionable. At a macro level the CAGR until 2012 is predicted to be of the order of 14% 97 Although mobile phone access is a competitive industry in many countries. Forecasts from Gartner. One estimate of the excess cost borne by Europe as a result of the mainframe monopoly of IBM is of the order of $48 billion over a 20 year period.

Banking. growth and human welfare on account of its diffusion into different sectors of the economy and sections of the society. Forecast: IT expenditure by verticals Utilities Transportation Retail Trade Process… Government Healthcare Financial Services Communications 0 2000 4000 6000 8000 10000 12000 2012 (Forecast) 2009 (Forecast) 2006 (Actual) US $ Millions IT expenditure includes expenditure on Hardware. 39 . J. Software.101 Although India lost out on the productivity benefits of ivity IT diffusion. See also Ashok economy Jhunjhunwala. the contribution of IT to the economy can be thought of as possessing two interrelated features – IT growth itself and growth as a result of IT diffusion. Insurance. Chart 7 below shows growth of IT expenditure by vertical until 2012 (also see annex XVI) XVI). September 2007. Internal services and IT services. However. 101 Information Technology and Productivity: Evidence from India’s Manufacturing Sector by K. Telecommunications. Joseph and Vinoj Abraham. Table 20: Sector Sector-Specific Policy for FDI in India Sectors Banks -Private Sector Insurance Companies Process Manufacturing Retail Trade * Services (Telecommunications) elecommunications) Transportation Source: Ministry of commerce 100 FDI Equity Cap 74% 26% 100% 51 % 74% 49% Year of liberalization 1990 1999 2006 1998 1991 Analytically. The former refers to the contribution in output. Retail etc were belatedly liberalized and Foreign Direct Investment (FDI) restrictions were eased at different points in time across sectors due to varying degrees of political economy pressures (See Table 20). diffusion of IT in India currently remains at low levels100. an unintended by product of being a late starter has been the ability to leapfrog legacy mainframe applications.and the corresponding dollar value of the market is estimated to be about 50 billion dollars. competitiveness. Academic research has arying cademic shown that IT investment has a positive and significant impact on both partial and total factor productivity in the adopting sector. employment and export earning arising from the production of IT related goods and services and the latter xport refers to IT induced development through enhanced productivity. IT intensive sectors like Telecommunications.

In addition.The server industry’s landscape has changed significantly in the last few years. competition developed to become the most intense. partly because these powerful data-serving computers often run “mission-critical” applications using proprietary operating systems.(Also see. just like other speciality engines designed to accelerate WebSphere middleware (zAAPs) and DB2 database routines (zIIPs)105. widespread use of technology.com/systems /z/?cm_re=masthead-_-products-_-sys-zseries and http://h20223. zAAP. they were confronted with a wider choice of vendors and technology across all the segments of the server market. 102 103 Interview with Intel India The power of the chip increases as the number of cores per chip increases. The emergence and popularity of the Linux platform for servers resulted in IBM slashing the prices it charges for its System z mainframe engines that are set up to exclusively run Linux. zIIP) are all exactly the same physical microprocessor as the standard z10 processor (which IBM calls a “general purpose processor”). Shrinking size of transistors has meant that 1.Gartner) 40 .www2. These are sold at much lower prices than standard mainframe engines. Last year Hewlett-Packard (HP). albeit in the entry level segment.com/NonStopComputing/cache/ 307953-0-0-0-121. The increasing power of the chip. the mid and high end servers proved a little more resistant to commoditization. In the early days of computing companies had to buy a mainframe for their high end IT requirements which cost millions of dollars or share one with someone else. The only difference between the specialty engines and the GPP is that the functionality of the specialty engines has been limited by small changes in the microcode which prevent the execution of legacy workloads on these engines.hp. based purely on the extent to which the applications running on these processor face or do not face high exit barriers. In this example of price discrimination. and standardization paved the way for “commoditization” a term most commonly applied to the PC industry. while the enterprise level segment also witnessed increased competition due to the emergence of HP and Sun in the category once dominated by IBM. Accessed from http://www-03. Servers also quickly commoditized. EPIC architecture based Itanium processors are now deployed as an alternate computing platform for both IBM P-Series RISC servers and the IBM Z series mainframes104. On the other hand. bought Electronic Data Systems (EDS). The IT industry has also benefitted from ‘Moores Law’. Rapid innovation has seen the power of the microprocessor clock a nine fold increase in performance over the last decade102. a big provider of computer services. See Annex XVII for Intel’s technology development timeline for the microprocessor. IBM is able to sell exactly the same product to the same customers at prices varying in a ratio of 1 to 10. The various IBM “specialty engines” (IFL. while quad core has become familiar jargon in the computing industry103. giving HP more manpower to help its customers build more advanced data centres and effectively compete with IBM in the entire value chain. the internet boom created additional demand for high-end computers providing space to accommodate multiple vendors and technologies. At the entry level.8 billion of these can now be tightly packed into a microprocessor.Specialty engines are key factors in IBM mainframe viability. the so-called Integrated Facility for Linux (IFL) speciality engines.html 105 IBM halves mainframe Linux engine prices Feeling the Nehalem Xeon pinch By Timothy Prickett Morgan Posted in Servers. Eight cores in a chip are in the development stage 104 HP Integrity Range of servers which competes with IBM P-Series RISC servers and HP Non Stop Server line up which competes with IBM Z-series mainframes. IBM may not have had a lot of direct competition in the enterprise IT market in the past but the story is different now.ibm. Likewise Oracle’s acquisition of Sun Microsystems will also create an integrated IT provider thereby matching IBM in the breadth of offerings. 17th August 2009. the world’s biggest computer-maker. When Indian companies began their purchase of servers in the late 1990s and early part of this century.

During our survey. PL/I. A break up of these by operating system is not available. The small set of Indian IBM mainframe customers running z/OS implies that the current ‘losses’ from lock in are likely to be low. Mainframe usage in India is much lower compared to China. Assembler etc) are on z/OS and these result in additional sales. Java application servers compete with Microsoft’s . Contrast this with system Z sales worldwide where vast majority of the sales are upgradations. WebSphere. In addition if one is forewarned about the dangers of lock in. even newer applications are being run on the z/OS in addition to legacy applications. Within new sales. The results are astonishing. is entitled to use the much less expensive zAAP specialty engines. One of the critical reasons cited against the mainframe was high price. we were told that if you were looking for an artificial heart. the deadweight losses could be higher like in the case of Europe cited earlier. In effect. The average price of the mainframe with z/OS is one and a half to ten times more expensive in China. Because of this. 107 For example. it would weigh heavily against IBMs z/OS. Linux is more popular than the z/OS.As a result of these developments.Net application server (which is very similar in conception but uses a slightly different language. would you go for the one with the lowest price? 108 A consultant had once famously remarked that IBM works best in an environment where the clients know little. albeit they do exist. the number of Z mainframe users in India is only 25. 41 . although none of the respondents including those who had opted for it were willing to share price information with us. The reason is mainframe with Linux provides cheaper computing. the price of the mainframe is comparable to the price of an HP Superdome 9000 series in India (See Table 21). Our survey however does reveal that the mainframe is an expensive system compared to the others. WebSphere is a piece of middleware that executes applications written in Java. According to IBM. On the other hand. Our primary survey did not provide any insights explaining why mainframes are more popular in China. C#). we found that two large Indian private sector banks were faced with a similar decision in the early 2000s for their credit card application. We were informally informed that about 60 per cent sales of the mainframe in India are due to legacy. In addition. the State Bank of India. If the installed base becomes larger in the future and the z/OS remains a proprietary standard. One of the banks chose to run the application on the Z mainframe with z/OS while the other chose a Unix environment on the HP Itanium server. Websphere is one of the best selling application platforms which is also available on z/OS. WebSphere competes with Oracle’s WebLogic and Red Hat’s JBoss (which are also Java application servers). A number of legacy applications (COBOL. IBM too did not disclose prices. chose the HP superdome hardware and HPUX as OS over alternative solutions including IBMs mainframe. India’s largest public sector bank. although in last 5-6 years mainframes with Linux are increasing. Note that even when running on z/OS. see section V above. What this implies is the dangers of lock in to the proprietary standard z/OS on the mainframe are likely to be low for India. Based on this sparse information one can conjecture that IBM. which is Java application server software. but a number of customers do look for hosting current generation applications on z/OS-Websphere combination106. We discuss this later in this section alongwith other developments likely to occur in India. More broadly. We therefore did a comparison of the average price of mainframes shipped with the z/OS in India and China from secondary data. while the rest are new applications. albeit also one with a perception of higher quality107. 106 An exact break up of z/OS and Linux on mainframe in India is not available-Interview with IBM. but it is growing. One of our respondents was of the view that users would be well advised to create a mainframe users club to exchange relevant information among them to create countervailing pressure on IBM108.

Table 21: Mainframe and Superdome’s average price in India and China Country India India India China India China India India India China India China India India India India China Source. On the other hand.e.467 1.194.128 4. Increasing commoditization of the entry level segment means that vendors in this space compete mainly on cost. and is significantly less concentrated than the high-end or midrange segments. The idea is that computing will increasingly be delivered as a service. This conjecture gains some credence if one considers the strategic contracts that IBM has inked with telecom companies such as Airtel.721 868. as we have shown results in greater lock in.557. Documents.466 732. HP and Sun.widely regarded as an expensive brand. the enterprise level segment has proven to be more resistant to commoditization.691 2.047.918 288.498.766 450.040.976 484.382. making 42 . Idea and Vodafone.253 449. e-mails and other data will be stored online.661 1.029 2008 2008 2008 2008 HP 9000 Superdome Superdome System z9 BC System z9 EC HP-UX HP-UX zOS/OS390 zOS/OS390 2009 2009 2009 2009 HP 9000 Superdome Superdome Superdome System z9 EC HP-UX HP-UX Linux z/OS.816.985.330 1.747 885. is reviving efforts to gain entry into the young Indian market even by ‘under pricing’ its products compared to other markets.044. or “in the cloud”. over the internet. A related finding is that the lower end of the market is dominated by x86 machines.963. create pressure for cost reductions as enterprises shift to services ‘in the cloud’. although interestingly the platform used by IBM is the P-series AIX instead of the Z mainframe and z/OS which.140 2.574 1.OS390 Our study clearly demonstrates that the high-end server market in India is highly concentrated and dominated by Unix machines from IBM.941 1.313 1. In all three cases. from vast warehouses of shared machines.Gartner Data Year 2007 2007 2007 2007 Brand HP 9000 Superdome Superdome System z9 BC System z9 EC OS HP-UX HP-UX zOS/OS390 zOS/OS390 Average price (US $) 1. IT infrastructure is wholly outsourced to IBM (See Airtel Case study in Annex XIII). A technological development called ‘cloud computing’ however could do to the enterprise segment market what commoditization has done to the entry level i.

Since IBM is the only vendor of IBM-compatible mainframes. While cloud computing is attractive as a concept. it is becoming increasingly rare to find IT staff with the necessary skills to administer mainframes. many risks are present. Sun and to an extent Cisco will dominate this segment. as it stands today.com/opinion/displaystory. The former two are integrated IT companies (and along with Oracle-Sun in the future) will increasingly dominate the enterprise server market in India. there are actions that regulators can take to address abuse by the monopoly provider under competition law. Many things work this way already. It is noteworthy that the Government of India (GoI) has in fact been advocating open source software.them accessible from any PC or mobile device109. If one cloud is not compatible with the others. However. In house IT support is of course of no use in case specific skills required for maintenance are lacking. for the next five years in India. given that there are still many businesses and government departments that have to rely on mainframe systems. Our study has shown that presence of a skilled in house IT team can go a long way in reducing the monopoly power of the vendor in certain cases. All the large public sector units and government departments we interviewed placed a high degree of reliance on their own IT departments. First is the well-known risk of technological lock-in as rival companies promote their own. the technology is still new and will take several years to mature and grow to be robust. Economist 15th Oct 2009 http://www. cfm?story_id=14644393 110 Massimano Passini. it is therefore a good idea for IT solutions to be based on open systems in an organization since service support is more easily available. Regulators need to be therefore cautioned about the existential dangers of such dominance extending to software and professional services. Government procurement in India uses a tender approach and has a life on average of about 5 years after which fresh bids are invited with no apparent benefit to the incumbent IT provider. mutually incompatible. standards and formats. IBM. Analyst Gartner 43 . As with any new and untested technology. is therefore unlikely to displace the dominance of the big three in the enterprise level segment for the next three or realistically. It is widely believed that open source software (OSS) can be considered as an alternative to the 109 Battle of the Clouds. it has a virtual monopoly is support and maintenance as well. Unless the application software is standardized and available off the shelf from another software provider. Cloud computing. As discussed in the report the lock in and the corresponding switch from an open standard is easier compared to a proprietary standard. A lower bound to this was put at three years by one expert we interviewed110. For example. This is discussed ahead.economist. Most IT professionals globally and in India are more focused on learning volume platforms based on open systems where their skills are more portable from job to job. a movement that is still in its infancy. Second is the risk to privacy and finally and most importantly for enterprises running mission critical applications is the risk to safety and security of their data. Other things the same. the integrated IT vendor has a comparative advantage in providing not only the software but also maintenance and support services to the client. Entry barriers in this segment are high and competition from cloud computing seems a long way off. the danger of lock-in is merely transferred from one technology to another. HP. as they have done in the past. Our survey suggests that large enterprises prefer to buy hardware bundled with the operating system. from e-mail and photo albums to calendars and shared documents.

were of the opinion that the TCO of OSS is no different from proprietary Unix based software. including the quality of 111 112 “Economic Impact of Free and Open Source software . a mandate to use only OSS could be a tool to address the monopoly of a proprietary OS. The study based on 20 case studies of Indian organizations drawn from government departments. At the same time only few states (Chattisgarh. 115 While government is willing to promote OSS and Open standard to lower entry barrier and hence increase competition. 44 . National Policy on ICT in school education gave preferential treatment to FOSS. In addition the larger organizations in our sample were unwilling to risk running mission critical applications on OSS112. interoperability is an important issue in formulating a standard113.proprietary software to reduce cost and increase transparency. all open standard is however not open source. OSS must not be confused with “free” software that is distributed free of cost. National Capital Territory of Delhi. It can be implemented by number of players or users for their own specific purpose. Without going into the merits of the methodology adopted by the study. but as yet there is no clear mandate or policy that directs use of OSS114. and Tamil Nadu) have a separate ICT policy which recommend use of OSS. However. Some claimed that OSS fails when number of transactions/instructions crosses a certain limit. No single individual or company controls the open standard.A Study in India” Rahul De. A recent study conducted by IIM Bangalore analysed the economic impact of FOSS (free and open source) in India111. IIM Bangalore.115 How effective this will be will depend on various factors. we cannot run an organization on somebody else’s passion. It is important however to distinguish between open source and open standard. hence casting doubts on its reliability and scalability. MICT came up with a draft policy on multiple open standards for e-governance project. Government deals are important for vendors not only for their own sake but also to bolster subsequent bids. the conclusion that FOSS will result in huge cost savings is a claim that needs to be treated with a measure of skepticism. OSS is also ‘free’ software in the sense of ‘free’ speech or ‘free’ ideas but not free of cost. Since both central and state governments in India are likely to generate huge demand for servers both for public enterprises and more importantly for citizen services. especially in the context of the market segment we studied for this report. Free software may or may not be open source. although it is not mandated in any of the States. Our respondents. Like other developing countries GoI encourages use of open source. we are running a business not a research organisation etc. Some firms were running OSS in parallel on trial basis but relied exclusively on the proprietary platform to run mission critical applications. One must keep in mind that while all open source is open standard. At the same time they are not willing to give up the functionality of proprietary software just by considering the lower cost of OSS. Open source software is that for such software any one can look at the source code. industry experts still prefer to rely on proprietary software to reduce the risk and uncertainty. A government diktat to use only OSS will be a big blow to the monopoly of proprietary systems like the z/OS. depending on whether or not the source code of the free software is made accessible to all. bar none. commercial firms and educational institutions. There was a fair amount of skepticism relating to OSS among the respondents that manifested itself in statements like. Therefore. in general an open standard is a format that is publicly available and has various rights to use associated with it.com/newsite/drupal/osvsos 114 Central governments have advised Indian Railway to use FOSS in their desktop and laptop. 113 http://sutor. Although there is no general consensus about the definition of open standard. Kerala. Conceptually the free nature of OSS allows users to customise software according to their need without moving away from an open standard.

industry experts estimate. patent claims etc. Competition Act. For example.co. For every rupee of IT spend on the UID project. Closed standards would create serious patent and interoperability complications. The Competition Act does not define any specific degree or percentage of market share that is relevant to be considered as important in the analysis to establish AoD. The Government has appointed Infosys founder and vice chairman Nandan Nilekani as the head of Unique Identification Authority of India (UID-AI) with an initial budget Rs 120 crores. An enterprise is said to be in a dominant position118 under the Competition Act when it is in a position of strength which enables it to: (a) (b) Operate independently of competitive forces prevailing in the relevant market. but the risk would be in ceding some control over the information to IBM as a result of the proprietary standard116. Biometrics (which includes fingerprint. The Act recognizes that such unilateral conduct can be exercised by enterprise acting on its own. besides systems integrators. storing large data sets and performing online verification on IBM mainframe z/OS will be eminently possible.given that it has successfully extracted huge profits from proprietary mainframe offerings for so long and if public action is insufficient to reign in the monopoly. It is unlikely that a company like IBM will change its business practices .000-20. Technical specifications of the project are not available but openness and interoperability are likely to be the basic requirements. The focus of analysis on IBM’s conduct should naturally be from the perspective of whether or not there is or has been abuse of dominance (AoD). or Affect its competitors or consumers or the relevant market in its favour. 15th December 2009 accessed from http://www. face and iris recognition) and computing power hold the keys to the UID project which is estimated to offer a Rs 15.theregister. underlining India’s potential demand for computing. Annex XVIII provides the framework within which any enterprise is investigated for AoD. Vital data of this nature should be stored in formats that are open and free of all constraints like royalties. The primary concern of the law therefore is with unilateral conduct abuses principles of competition. In this study we have shown that the high end market is dominated by the big three. database. or by two or more enterprises acting as a ‘group’.000 crore opportunity to computing. it will still be a small percentage of the total demand for IT from Government over the next few years. While the UID project will generate massive demand for IT. around 60 per cent of the spending will go to hardware vendors.OSS.uk/2009/12/15/neon_zprime_ibm_lawsuit/ 118 Explanation to Section 4(2). The approach of EC. the only alternative course of action is under competition law117. The first set of unique identity numbers will be issued in the next 12-18 months and the UIDAI plans to cover 600 million people within four years. Canada and the US on the issue of market share could be instructive in this regard and is summarized below for reference: 116 Financial institutions in the US on the z/OS have been locked in to the legacy application for many years at high cost 117 Neon sues IBM over 'anticompetitive' mainframe tactics. 45 . This issue has already come up for debate in India in the context of the Unique Identifier (UID) project which will create huge demand for server’s hardware software and applications. smartcard and storage vendors.

18 (10th Cir. However. are not likely to enjoy a (single) dominant position on the market concerned. competition authorities in other jurisdictions have found themselves in positions whereby a finding on a ‘dominant position’ is supported by evidence relating to its abuse. The holistic analysis will necessarily include size and resources of IBM and its competitors. and relying on EC’s and Canada’s approaches. These concepts cannot therefore be looked strictly as distinct elements. in several situations. although some undertakings with market shares below 40% could be considered to be in a dominant position. the analysis of the Commission is as follows: “It is very likely that very high market shares. provided the rivals hold a much smaller market share of the market. European Commission. and the EU. Courts have held that a market share below 50% precludes finding monopoly power120.). December 2005. This would be the case where an undertaking holds 50% or more of the market. and the leading treatise suggests that a share of over 7075% for at least five years is required. etc. 1984) (— monopolization is rarely found when the defendant‘s share of the relevant market is below 70%“). 120 See. economic power of IBM including commercial advantages over competitors (including technological lead.3d 1406. 1995) (Posner. 1989) (To establish —monopoly power.123 119 DG Competition Discussion Paper on the Application of Article 82 of the Treaty to Exclusionary Abuses. market share in itself is only one of the factors for assessment of dominance.”119 In Canada. However. undertakings with market shares no more than 25%.• • • In the EC. indicate a dominant position. Natural Gas Pipeline Co. It is important to note here that there are no strict demarcations as to when an inquiry into a ‘dominant position’ gets over and analysis of whether there is an ‘abuse’ begins. of America.g. 694 n. In the United States. than below 40%. 748 F. 121 See Colorado Interstate Gas Co. Exxon Corp. vertical integration of IBM and of course market structure and size of market. an arguable case could be made that the market share of IBM in India is sufficient to warrant a closer scrutiny since it clearly has the largest market share. 65 F. which have been held for some time. India’s Competition Act draws on principles from competition law in the U. But a market share of 35 percent or more will generally prompt further examination.) (—Fifty percent is below any accepted benchmark for inferring monopoly power from market share. 122 See section V 123 Section 27(a) of the Competition Act 46 . If therefore the CCI finds that there exists a case of abuse of dominant position it may pass all or any of the following orders: Order for discontinuation of the abuse of dominant position. a recent Discussion Paper on Article 82 (which deals with the subject of abuse of dominance). lower courts generally require a minimum market share of between 70% and 80%.. its IPRs.121 Based on the above assessment.S. 1411 (7th Cir. dominance is more likely to be found in the market share range of 40-50%. rather they would constitute part of a holistic analysis. In fact.2d 937. Blue Cross & Blue Shield United of Wisconsin v. Berwick Bay Real Estate Partners. e. the Competition Bureau’s approach has been as follows that a market share of less than 35 percent will generally not give rise to concerns of market power or dominance. Marshfield Clinic. 940 (5th Cir. v. In case of lower market shares.2d 683. 885 F. v.“). J. Further nuance in defining the relevant market by workload type (new versus legacy) gives IBM almost certain structural dominance in the market122.

In addition to this.127 Theoretically. Clearly that is not the case today.e.126 The scope of the orders italicized in the last two points above.124 Direct the enterprises concerned to abide by such other orders as the CCI may pass and comply with directions including payment of costs if any. potential remedies include orders relating to: (i) unbundling of software from the mainframe server. and in particular in the mainframe and high-end server markets. Because the Indian enterprise IT market is entering its high growth phase. (iii) licensing of IBM’s software on terms that are determined to be reasonable and non-discriminatory. the manifesto points out. customers should be able to move their data and applications easily from one to another. It would indeed be ideal if the cloud manifesto’s principle that data and applications should be allowed to move freely were applied everywhere in enterprise IT. Several instances of IBMs abuse of its monopoly position such as high cost and stifling innovation are well known (See Annex XIV). a cloud manifesto has been released. CCI has an excellent opportunity to avoid the pitfalls into which the more mature IT markets have fallen. indicate the vast nature of powers that the CCI potentially has in designing remedies specifically to address the specific case at hand. due to various reasons. IBM has a considerable stake in the new technology known as “cloud computing”. On the other hand in India. and there is a very striking contrast between the principles stated in the manifesto and the principles that IBM practices in the mainframe market. and any other remedy that will ensure competition in the market. (ii) actions necessary to ensure interoperability. In this regard. In the US and Europe. it would be possible for the CCI to design remedies appropriate to each set of circumstances.125 Pass any such other order or issue directions it may deem fit. the CCI is also empowered to direct division of an enterprise enjoying dominant position to ensure that the enterprise does not abuse such dominant position. IBM is alleged to have successfully extracted huge profits from proprietary mainframe offerings for long and it is unlikely that this will change in the near future. As we stated above in the report old habits die hard. the crucial server operating systems) is the first policy recommendation. there are no indications of IBM having conducted business as it has in the West. That is no assurance however it is precluded for all times to come. 124 125 Section 27(b) of the Competition Act Section 27(e) of the Competition Act 126 Section 27(g) of the Competition Act 127 Section 28(1) of the Competition Act 47 . Since there will be many different computing clouds. A format war is likely to be played out over the next few years in this domain and to nip any monopolisation effort in the bud.Order of penalty of not more than 10 percent of the average turnover for the last 3 financial years upon each person/ enterprise which is a party to the agreement or abuse. The “manifesto” essentially calls for computing firms not to fall back on bad old habits by trying to lock in customers. Tying operating systems to a particular brand of hardware is the fundamental mechanism by which all the high-end enterprise system vendors – not just IBM – seek to lock in their customers and raise switching costs. to require that the sale of enterprise class server hardware shall not be tied to the sale of enterprise software (in particular. Unbundling hardware and software i. Therefore steps should be taken to ensure that this market remains open to multiple suppliers and platforms. therefore. One of the main players behind the manifesto is IBM.

and rather significant even in the short term. or Linux. This is our second policy recommendation. Windows Server and to some extent Sun’s Solaris (which runs on Intel or AMD servers as well as on Sun’s proprietary SPARC servers. Customer choice can be enhanced if server operating system vendors (all of them. Interoperability is a more complex issue but it is partially addressed in the two policy recommendations above. and European regulators with respect to IBM as pointed out earlier in the report. customers should be allowed to make this choice. In fact. on Intel or AMD servers using emulators such as TurboHercules). Customers should be allowed to choose z/OS (or HP-UX. the consequences of IBM’s tying of z/OS to its own mainframe hardware (to the exclusion of would-be competitors seeking to offer IBM mainframe emulation on Intel servers) have been somewhat limited so far. a competitive IT sector will confer substantial direct and indirect benefits to the India economy. or Windows Server. India would be much better off if CCI is able to successfully uncouple the lock-in strategy of the large server vendors before they come to completely control the market. For reasons already documented in this report. in fact this would merely be a return to the decadeslong practice of U. these are likely to be large indeed in the long term. etc. IBM itself has also recently introduced such a mainframe emulator. But vendors should not be in a position to force customers to purchase only their brand of server hardware once the choice of server operating system has been made.) if they wish. This is nothing new. If customers are allowed to purchase the server hardware of their choice for use with the server operating system of their choice (unbundled hardware and software) there may be nothing wrong with the fact that z/OS is not compatible with more modern operating systems.S. Requiring IBM to reveal its source code a la Microsoft is not going to be a pragmatic solution. It could even be the case for z/OS. 48 . It would probably be much healthier for Indian IT and for the Indian economy if enterprises developed their complex business applications on server operating systems that are not tied to a particular brand of hardware. However.g. because mature mainframe emulation software products have existed for a number of years which allow z/OS to run on Intel server hardware. the report also points out that high-end server market as a whole in India is highly concentrated. It could theoretically be the case with HP-UX (HP version of Unix).The report finds that since there are few z/OS mainframes in India. so we refrain from making it. Today this is obviously the case with Linux (open source). not just IBM) are required to license their software on RAND terms (Reasonable and Non-Discriminatory) These terms however would in no way restrict the right of vendors to earn fair revenue and profits from the sale of their proprietary software. Such products include FLEX and TurboHercules. but limits its use to software developers only and refuses to sell it to production customers. since HP’s SuperDome servers are based on Itanium processors. on the basis of their own criteria of performance and functionality. Although we cannot quantify the economic benefits to India of such “pre-emptive” action. To the extent that there is no inherent technical barrier to running z/OS on non-IBM hardware (e. India has an opportunity to leapfrog the more mature IT using economies by adopting a forward-looking regulatory stances that prevents IT system vendor lock-in from establishing itself as the norm in the Indian economy. the leading Unix vendors such as IBM and HP do compel users of their Unix operating systems to purchase their brand of server hardware.

Bresnahan T. Constance E.org/opensrc.M. "Open-source vs. Thomas Hazlett. 21. Steinberg James (2003). N. Vol. pp. “Mainframe Migration Case Studies: A Total Cost of Ownership Comparison” An Alinean White Paper. 3. 643-681 13. “Information Technology & Development: Beyond “Either/Or”. Free Press. http://cds. 7.” Harvard Business School note 9. “The Antitrust Paradox”. Information and Communication. “Legal and Economic Aspects of the Microsoft Antitrust Case in the Information Age” Business Economics. Schmalensee Richard (1982). (2001). Garrouste (eds.pdf Books 1.pdf 10.. 65 No. Gould Jeff (2004). Garzarelli.pdf 2. Jullien. Available at: http://openmainframe. “ Open Source Software: Free provision of Complex Public Goods”. No.nabe.” chapter in J.edu/download_files/wp389.Bibliography Papers 1. vol. Greg Shanker (2008).1. pages 47-57 6. Robert Litan. Bessen James (2005). Dalle J. “Note on Application of the Antitrust Laws to the New Economy: An Analysis of United States v. New York 2. Working Paper series in Boston University .peerstone. Bagley (2001). Free/Open Source Research Community. 49 . “Economics of Information Technology” University of California. Varian Hal R. “Antitrust and the New Industrial Economics”.School of Law Available at: http://www.pdf 5. Vol. “Linux and ERP: A White Paper” A White Paper by Peerstone Research Available at: http://www. "General purpose technologies engines of growth?” Journal of Econometrics. (1995). M.org/downloads/resources/ 9. “Industrial Economics: An Overview”. Abraham Vinoj (2007). Trivandrum. “Information Technology and Productivity: Evidence from India’s Manufacturing Sector” Working Paper Series 389. Jeffrey Robert Church and Roger Ware (2000). Kerala. Schmalensee Richard (1988). 2 (Spring. American.com/publib/be/000245. Microsoft Corporation. pp. Joseph J.mit.researchoninnovation. Berkeley 3. “Benefits of Mainframe Competition for the European Economy”. Demsetz Harold (1982). Bork H. MIT. The Brookings Review. Gould Jeff (2009).edu/ papers/dalle2. The Economic Journal.pdf 11. Vol. 8. 98. proprietary software". 24-28 12. “Industrial Organization: A Strategic Approach” The McGraw-Hill Companies Inc. (2002).com/pdfs/Peerstone_Linux_ERP_ White_Paper. pp. Trajtenberg. 72(1). 45-48 Published by: The Brookings Institution 14.. Centre for Development Studies.). 2003). No. Robert (1978).802-090 4. American Economic Association. "Barriers to Entry." American Economic Review. and Edwin Rockefeller (2000). Markets. available at: http://opensource. Birner and P. 72 (2)..83-108. Giampaolo (2004) “Open-source Software and the Economics of Organization. Available at: http://www. 392. Vol. K. Economic Review. London and New York: Routledge: 47-62. pp.

“National Policy on Open Standards for e-Governance”. “Dataquest Guide: Software Market Research Definitions. Ovum (2005). Microsoft Whitepaper (2003). 2009). John R.org (2009). “Fujitsu OpenVME costs compared to IBM mainframe costs”. 2006). Barnett Gary 12. “India 2009 IT Market Predictions”. 2009). “Dataquest Insight: The Realities and Challenges of Linux in Brazil. Carl Claunch (Sep 24. Integration Systems LLC (2009). Independent Analyst Report by Graham J Barry http://www. 0. Smart Computing and Xephon Plc (2002). “Understanding IBMs Mainframe Monopoly” 11. “Study on the Domestic Services (IT -ITES) Market Opportunity” 8. 4Q08 and 2008” 6. “World Telecom Development Report” 5. “Technology review of mainframe computer systems and their alternatives”. ITU (2006). 2008).And More ROI -2008 to 2014” Gartner Research Reports 1. done by a team of Rahul De at Indian Institute of ManagementBangalore 9. http://www.pdf. “Specialty Engines are Key Factors in IBM Mainframe Viability” 50 . Worldwide. Free Press (1985). Adrian O’Connell (March 25. Colleen Graham (Feb 23 2009). A modern assessment by Terry Keene 6. OpenMainframe. Jennifer Wu (Oct 16. 2009). 4. 2009). “Dataquest Insight: Server Forecast Assumptions. Worldwide. “Economic Impact of Free and Open Source software A Study in India”. Springboard Research document (2009). Aman Munglani (July 25. Heeral Kota (Oct 15.org(2008). “Mainframe linux Benchmark Project” 7. 2009). “Key Issues for Server Markets” 2. Government of Kerala. 2009-2014” 11. Ltd. Worldwide. “Market Trends: Servers. March 2008” 5.0 by Ministry of Communications & Information Technology 3. Heeral Kota (April 20. Draft national Policy version 2. Wintergreen Research (2008). 2009). NASSCOM-IDC (2006).org/web/main/ITPolicy-2007. “Dataquest Insight: Indian Storage Market Moving Away From Dependence on Server Sales” 3.fujitsu.com/downloads/ZA/whitepapers/openvme. Heeral Kota (April 17. 1Q09 (Preliminary Statistics)” 7. “The future of the mainframe”. “Dataquest Insight: What to Expect in Asia/Pacific's IT Services Landscape From HP's Acquisition of EDS” 9. “Market Trends: Servers. “Mainframe As A Green Machine -. E. Manish Bahl 14. “The z10 is Not Your Father’s Mainframe”. Asia/Pacific. “Information Technology Policy: Towards an inclusive knowledge society”. 2. “How Cloud Computing Relates to Grid Computing” 4. Jacqueline Heng (Dec 10. 2008). Red Hat India Pvt. Tichy F. OpenMainframe. “Quarterly Statistics: Server Shipments.keralaitmission. Phelps (August 18.pdf 13. “The Economic Institutions of Capitalism”. China and India” 10. 2Q09” 8. 2009). Jennifer Wu (Feb 10. Williamson. Department of Information Technology (2007). Walter 10.Reports 1. (2009). Department of Information Technology (2009).

Jeff Goldberg. 2009).. “Bharti Airtel grows at a stunning pace by keeping its focus on the customer”. Available at:http://www. 2008 and Beyond” 18. Available at: http://tech2. Worldwide” 16. “Much ado about IBM's mainframe monopoly” Timothy Prickett Morgan. Naveen Mishra (July 22. 2009 4. Jasmine Desai. Biztech2. Feeling the Nehalem Xeon pinch”. 2009 5. Thomas J.html?ERL=true&jumpid= reg_R1002_USEN Accessed on May 14. Available at: http://insuranceblog. Lillian M. 2009 3. Thomas J. 2009). 2009 6.co. “Mainframe: The Dinosaur That Is Not Quite Extinct”. 2008). “Getting Starting With Private Cloud: Services First” 21. Uko Tian (June 15. 2008).12. Available at: http://www. “User Survey Analysis: Are Linux Servers a Strategic Choice or TCO Reduction?” 14. 2008). “Mainframe rising”.com/biz/india/features/computing/ mainframe-the-dinosaur-that-is-not-quite-extinct/47602/0 Accessed on July 23.uk/2009/10/20/ibm_mainframe_i_monopoly/ Accessed on Oct 23. “Market Trends: Servers.com/software/ success/cssdb.hp. Laurie F.com/2009/10/is-there-such-a-thing-asa-mainframe-monopoly/ Accessed on Oct 15. 2009). 2009). 2008).in. Alvarado (Dec 17. “Quarterly Statistics: Servers by Vertical Market and Size of Business. Express Computer (April 13. “2008 Data Center Conference Poll Findings: The IBM Mainframe Market” 17. 2009-2013” 15.ibm. Celent Blog (Oct 15. “Is There Such a Thing as a Mainframe Monopoly?”. “The Real Story about the IBM Mainframe Makeover”. Mike Chuba (March 20. India. 2009). 2009). 2009). “Market Trends: Application Development Slow but Steady. Richard Ni (March 6. 1Q09” Articles /Web Blog /News items 1. “SWOT: IBM Servers. Akhtar Pasha. 2009). Available at: http://www. Hewlett-Packard Development Company (Feb 20. 2009).com/20090413/market01. “IBM halves mainframe Linux engine prices. Channel Register (Aug 17. “Emerging Market Analysis: IT. Kiyomi Yamada (March 17.expresscomputeronline.www7. 2009 51 . Wurster (14 Oct 2009). 2009).co.channelregister.com/ERC/cache/ 582432-0-0-0-121. Worldwide” 13.channelregister. Available at: http://www-01. 2009 2.celent. Channel Register (Oct 20. 2009). Worldwide. Asia/Pacific. “Server Virtualization: One Path That Leads to Cloud Computing” 22. 2008). Bittman (Oct 29. Kiyomi Yamada (April 9. Bittman (Oct 22. IBM Corporation (March 11. “Q&A: The Many Aspects of Private Cloud Computing” 20.shtml Accessed on June 16. Timothy Prickett Morgan.uk/2009/08/17/ibm_mainframe_linux_cuts Accessed on Aug 17. Bittman (Oct 29. Available at: http://h71028. 2009 7. Thomas J.nsf/CS/JSTS-7C6M6P?OpenDocument&Site=gicss67tele&cty=en_us Accessed on July 21. 2009). “Singapore Leads Asia in 'Whole Government' IT Procurement” 19.com India (Feb 13.

com/article/technology-media-telco-SP/idINN09547181 20091009?sp=true Accessed on Oct 12. 2009 10.org (Aug 12.com (Oct 27. 2009). 2009 13. NetMediaEurope (Oct 16. 2009 17.html Accessed on July 5.html Accessed on Nov 4.com (Oct 2.infoworld. Available at: http://www. 2009 14.ppt Accessed on Oct 5. PC World.cfm?story_id=14258678 Accessed on Aug 17. “How to make a killing in the mainframe market”.com/businessfinance/economicsfocus/displaystory. Robert Hunt. Grant Gross. Available at: http://www. 2009 12. “IBM faces DOJ antitrust inquiry on mainframes”. 2004).eco nomist.pdf Accessed on Sep 25. 2009 11. The Economist (Aug 20. “IBM Mainframe Monopoly Probe Questioned By Analyst”. World's Largest Centralized Core Processing Implementation”. 2009). Available at: http://www. Tom Jowitt. Available at: http://www. for a Price”. 2008).main framemigration. “Standards. Available at: http://in.com/2009/10/ibm-profit-rises-despite-fallingsales.html Accessed on Oct 8.html Accessed on Sep 12. “Survey Revives Debate Over Mainframe's Future”.co. “Mainframe Sales Increasing in Southeast Asia.org/featuredarticles/ibm-says-998-of-mainframe-market-not-enough-we-want-it-all.tcs. “All Hail the IBM Mainframe: PSI Genuflects. Jeff Gould. 2009 20. 2009). Open Mainframe.org/featured-articles/how-to-make-akilling-in-the-mainframe-market.org/featured-articles/standards-openstandards-and-double-standards. Info-tecq.com/articles/?p=107 Accessed on June 4. Available at: http://www. Joe Clabby Available at: http://www.8. PC World.pdf Accessed on Oct 5. 2009).com/article/148042/all_hail_the_ibm_ mainframe_psi_genuflects_for_a_price. Available at: http://openmainframe.org “The closed cloud manifesto” Available at: http://openmainframe.com/businesscenter/article/ 174459/survey_revives_debate_over_mainframes_future. Open Mainframe. 2009 52 .org (Sep 24. 2008). IT Jungle (Dec 2.reuters. James Niccolai. 2009 22. 2008 ).info-tecq.org (Oct 2. “The Unix Guardian--IBM's p5 595 Tops the TPC-C Charts”. Open Mainframe. Available at: http://www. “Brain hunter Migration & Modernization Presentation”. we want it all” Jeff Gould Available at: http://openmainframe. 2009). Jeff Gould. Available at: http://www.html Accessed on Sep 3. Tata Consultancy Services (Feb.pcworld.com/d/hardware/ibm-faces-doj-antitrustinquiry-mainframes-118 Accessed on Oct 16.org/downloads/resources/closed_cloud_manifesto.org/brainhunter/Lists/Presentations/Attachments/2/Migration_Modernizat ion_Overview. Mainframe Migration Alliance (Sep 30.itjungle. “The Unkindest Cuts”.com (Oct 8. 2009).com/tug/tug120204story02. 2008).eweekeurope. PC World. 2009 15. Reuters India (Oct 9 2009).pcworld. open standards and double standards”. “IBM says 99.html Accessed on June 16.8% of mainframe market not enough. 2009). 2009 9. 2009 18.“Case Study: State Bank of India. Open Mainframe. Timothy Prickett Morgan.com (July 8. “ANALYSIS-IBM faces rivalry despite antitrust inquiry”.mainframeexec. Mainframe Executive (March 19. 2009 21. 2009 16. Available at: http://www.html Accessed on Oct 7. India and China”.com/sitecollectiondocuments/case%20studies/bancs_case_sbi. Stephen Haas. Available at: http://www. “IBM profit rises despite falling sales as margins”. Available at: http://openmainframe. 2009). 2009 19.uk/news/ibm-mainframemonopoly-probe-questioned-by-analyst-2120 Accessed on Oct 26.

“Intel and AMD: Agreeing on a pay-off”. Available at: http://openenterprisenews.openmainframe. “Battle of the Clouds”.economist.com www.com/printedition/PrinterFriendly.economist. 2009 24. “The mainframe isn't dead after all”.com/opinion/displaystory. The Economist (Oct 7. SUN. Available at: http://www. The Economist (Oct 15. 2009 27. HP. 2009).com http://www. 2000). 2009 Website www.economist.cfm?story_id=14644393 Accessed on Oct 26.gartner. “The new enforcers”. CISCO. “Unisys squeezes profits from revenue decline in Q3”.) 53 .top500. The Economist ( Nov 12. The Register ( Oct 28. Intel etc.tecn-news. 2008 ). The Open Enterprise (Aug 21. Available at: http://www.co.html Accessed on May 14.org/ Websites of IT companies (mainly IBM.org www.uk/ 2009/10/28/unisys_q3_2009_numbers/ Accessed on Oct 29. Timothy Prickett Morgan.cfm?Story_ID=387757 Accessed on Oct 7. 2009).23. Available at: http://www.com/businessfinance/displaystory. 2009 26. Available at: http://www. Jeff Gould.com/analysis/the-mainframe-isnt-dead-afterall.theregister. 2009 25. 2009).cfm?story_id=14885938&fsrc=r ss Accessed on Nov 17.

Middleware and Application software. In short. A database server is a computer system that processes database queries. Depending on the shape. Middleware sits "in the middle" between application software that may be working on different operating systems. but typically does not directly apply them in the performance of tasks that benefit the user. It hosts the several applications that run on a computer and handles the operations of computer hardware. Server hardware is the physical component of the system on which a set of instructions namely software is installed. 54 . Users and application programs access the services offered by the operating systems. An operating system is a software component of a computer system that is responsible for the management of various activities of the computer and the sharing of computer resources. transaction monitors. telecommunications software. Over the period of time technology advancement has makes the existing hardware obsolete. which is involved in integrating a computer's various capabilities. In other words. For the purpose of the study and for the purpose of defining the relevant market we have used the following definitions. meaning that they perform no other tasks besides their server tasks. A print server is a computer that manages one or more printers. The current focus now is to reduce the downtime and increase the productivity of the system. For example. application servers. Any user on the network can store files on the server. media players and database applications. a file server is a computer and storage device dedicated to storing files. size and architecture. and a network server is a computer that manages network traffic. which are used most often to support and simplify complex. it is the subclass of computer software that employs the capabilities of a computer directly and thoroughly to a task that the user wishes to perform. Users interact with operating systems through Command Line Interfaces (CLIs) or Graphical User Interfaces known as GUIs. This should be contrasted with system software (infrastructure) or middleware (computer services/ processes integrators). with the purpose of supporting or improving the software user's work. The software consists of a set of services that allows multiple processes running on one or more machines to interact. by means of system calls and application programming interfaces. except that it is stretched across multiple systems or applications. Application software is a computer program that functions and is operated by means of a computer. This technology evolved to provide for interoperability in support of the move to coherent distributed architectures.Annex I Definitions In the study we have analyzed the server hardware market. spreadsheets. Examples include EAI software. Middleware is computer software that connects software components or applications. and similar tools that support application development and delivery. It is similar to the middle layer of three-tier single system architecture. Servers are often dedicated. its corresponding Operating System (OS). and messaging-and-queuing software. It includes web servers. Typical examples of 'software applications' are word processors. servers are classified into several groups. In this context the term application refers to both the application software and its implementation. distributed applications. operating system enables user interaction with computer systems by acting as an interface between users or application programs and the computer hardware. A server is a computer or device on a network that manages network resources.

Annex II List of Players in the Indian Server Market Vendor IBM Hewlett-Packard Sun Microsystems Dell Acer HCL Wipro Fujitsu Apple Computer Lenovo SGI Stratus Computer Unisys Verari Systems Zenith Computer TCS Satyam Info tech Groupe Bull Hardware Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes No No Yes O/S Yes Yes Yes No No No No No Yes No Yes Yes Yes Yes No No No No Middleware Yes Yes Yes No No Yes No Yes No No No Yes Yes Yes No No No No Application Software Yes No Yes No No No Yes Yes Yes No No Yes Yes No No Yes Yes No Maintenance Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes 55 .

Examples would include Microsoft Exchange and Zimbra. In other words. E-mail/messaging — This category is composed of workloads designed for the handling of electronic mail and messaging. crash simulation and seismic analysis software. Data warehouse — This workload is a repository of an organization's electronically stored data that is designed to facilitate reporting and analysis through business intelligence tools. application development. process and middleware (PPMW) software. These workloads perform some manipulation of information through a client interface accessed via the Web as opposed to merely providing static Web pages to clients on a network. Front-end Web servers — These workloads provide HTTP responses in addition to optional data contents. This category is composed of software primarily for use by IT professionals. Examples include Apache and Microsoft Internet Information Services (IIS). Examples include Clipstream Video 3 and Wowza Media Server Pro. run and manage the performance of IT resources.Annex III Server Functions • Web applications — These represent the applications that are typically run in the second tier of a three-tiered server architecture. these workloads divide different parts of a given dataset and process those pieces separately and simultaneously on different servers to ultimately provide the individual results to form a larger. HPC — This workload is made up of parallel applications. typically in the form of Web pages. Streaming media — Servers running this workload provide audio and/or video playback to end users over a network. Infrastructure (excluding database management systems [DBMSs]) — The focus of infrastructure software is to build. such as HTML documents and linked images and/or objects. combined result. management. An example would include a shopping cart application accessed through a retail Web site. OLTP DBMS — This category represents OLTP database workloads. and portal. and examples include security. The focus of this workload is to process transactions that are linked to a relational database. • • • • • • • 56 . Examples would include digital content creation.

Annex IV Evolution of IBM in India 57 .

020 77.557 97.297 High-End Segment 2002 2003 2004 2005 2006 2007 2008 2009 Q1 0.250 0.307 77.429 Table 2: C4 for Asia Pacific region (APAC) by Vendor Revenue Years Total market Entry Level segment 77.423 0.895 76.465 0.162 0.313 0.856 98.430 74.148 98.336 0.262 0.505 98.293 78.265 0.577 78.178 0.980 97.379 0.269 0.322 97.402 0.348 0.697 96.173 0.196 0.400 89.892 99.338 0.306 0.532 98.082 98.837 78.330 0.916 99.265 98.989 80.319 0.978 90.596 99.172 0.919 78.312 99.178 0.223 0.873 90.889 99.422 96.096 77.995 98.187 0.132 High-End Segment 2002 2003 2004 2005 2006 2007 2008 2009 Q1 74.911 Mid range Segment 97.201 99.041 98.205 86.259 0.307 0.721 97.312 0.921 73.162 0.330 0.437 0.781 77.251 0.837 87.955 84.263 85.215 Mid range Segment 0.523 73.145 78.331 0.Annex V Estimates of HHI and C4 Asia Pacific Region Table 1: HHI for Asia Pacific region (APAC) by Vendor Revenue Years Total market Entry Level segment 0.204 0.192 0.384 0.389 0.304 76.462 98.196 0.198 0.166 0.042 97.380 0.358 99.209 0.693 87.426 0.187 0.226 0.334 99.318 0.736 58 .259 0.140 97.327 0.481 98.320 0.189 Mid range Segment 0.678 74.337 0.431 0.055 96.341 0.482 84.275 0.347 0.203 0.189 0.172 0.597 76.717 Mid range Segment 97.165 0.788 97.219 97.279 High-End Segment 2002 2003 2004 2005 2006 2007 2008 2009 Q1 0.379 0.380 78.800 Table 3: HHI for Asia Pacific region (APAC) by Shipments Years Total market Entry Level segment 0.207 97.013 High-End Segment 2002 2003 2004 2005 2006 2007 2008 2009 Q1 87.129 87.172 0.381 Table 4: C4 for Asia Pacific region (APAC) by Shipments Years Total market Entry Level segment 74.

222 0.85 89.02 85.87 83.185 Mid range Segment 0.182 0.00 71.33 85.54 75.00 99.00 99.79 100.384 0.233 0.234 0.21 85.223 0.48 74.52 80.07 80.447 0.00 99.338 0.00 99.343 0.204 0.198 0.330 0.54 92.246 0.363 Total market 2002 2003 2004 2005 2006 2007 2008 2009 Q1 0.181 0.60 Mid range Segment 100.66 86.03 73.205 0.00 99.188 0.00 100.217 0.00 85.79 100.349 0.22 77.89 Mid range Segment 100.168 0.86 90.381 0.42 85.359 0.93 100.201 Mid range Segment 0.406 0.226 0.88 100.53 83.183 0.339 0.257 Table 6: C4 for India by Vendor Revenue Years Entry Level segment 73.00 100.347 0.182 0.168 0.197 0.00 100.91 79.00 100.264 0.363 0.332 0.224 0.376 0.87 82.85 100.340 0.335 0.59 89.337 0.00 100.92 83.00 99.336 0.382 0.57 89.00 99.334 0.00 100.00 100.12 79.00 99.00 100.84 82.169 0.419 0.187 Table 8: C4 for India by Shipments Years Entry Level segment 71.46 100.347 0.94 75.76 59 .397 High-End Segment Total market 2002 2003 2004 2005 2006 2007 2008 2009 Q1 0.429 High-End Segment 0.337 0.66 100.88 87.358 0.70 80.281 0.80 Table 7: HHI for India by Shipments Years Entry Level segment 0.81 High-End Segment Total market 2002 2003 2004 2005 2006 2007 2008 2009 Q1 100.428 0.201 0.328 0.India Table 5: HHI for India by Vendor Revenue Years Entry Level segment 0.261 0.189 0.222 0.71 High-End Segment Total market 2002 2003 2004 2005 2006 2007 2008 2009 Q1 100.06 80.96 100.350 0.00 100.00 100.168 0.00 99.24 79.529 0.331 0.277 0.00 99.

532 99.259 92.590 High-End Segment 2002 2003 2004 2005 2006 2007 2008 2009 Q1 0.298 100.202 0.836 99.351 0.255 0.934 85.420 99.346 0.210 0.923 88.739 100.209 0.387 0.961 100.292 0.165 99.930 Mid range Segment 98.600 87.327 Mid range Segment 99.765 100.925 95.757 99.348 0.393 0.353 0.Australia Table 9: HHI for Australia by Vendor Revenue Years Total market Entry Level segment 0.349 0.650 99.227 0.026 80.242 0.777 86.751 82.934 94.367 0.358 0.248 0.000 99.361 98.601 85.512 99.232 86.244 0.423 86.540 98.807 100.431 87.250 0.000 High-End Segment 2002 2003 2004 2005 2006 2007 2008 2009 Q1 84.898 82.385 99.691 91.245 0.214 Mid range Segment 0.392 0.363 99.239 0.179 99.387 0.767 98.408 0.405 0.592 High-End Segment 2002 2003 2004 2005 2006 2007 2008 2009 Q1 0.223 0.243 0.812 92.000 High-End Segment 2002 2003 2004 2005 2006 2007 2008 2009 Q1 95.342 0.370 0.227 99.493 94.212 0.400 0.754 92.478 0.333 0.000 99.343 0.406 0.213 0.000 Table 11: HHI for Australia by Shipments Years Total market Entry Level segment 0.107 89.530 99.191 92.387 0.574 93.732 89.343 81.399 98.357 0.386 Table 10: C4 for Australia by Vendor Revenue Years Total market Entry Level segment 93.286 0.232 0.839 92.324 0.863 99.229 0.253 0.240 Mid range Segment 0.583 80.282 0.355 0.202 0.210 0.346 0.777 100.972 83.408 98.321 Table 12: C4 for Australia by Shipments Years Total market Entry Level segment 84.204 83.540 97.349 0.276 0.347 0.460 99.270 0.283 0.302 0.371 0.000 60 .353 0.262 0.430 87.670 99.244 0.854 99.

552 74.838 78.336 99.322 88.087 82.919 61 .397 0.269 0.018 98.218 Mid range Segment 0.937 99.257 0.351 75.269 High-End Segment 2002 2003 2004 2005 2006 2007 2008 2009 Q1 0.376 0.309 99.181 0.617 99.171 0.152 0.186 0.167 0.252 0.362 0.838 84.492 0.413 99.296 High-End Segment 2002 2003 2004 2005 2006 2007 2008 2009 Q1 0.941 99.419 0.China Table 13: HHI for China by Vendor Revenue Years Total market Entry Level segment 0.839 69.896 78.349 0.290 75.302 98.152 0.153 0.607 70.170 98.153 0.441 0.794 84.251 0.868 99.549 Mid range Segment 97.900 Table 15: HHI for China by Shipments Years Total market Entry Level segment 0.538 99.209 0.149 0.148 0.673 98.148 0.151 0.334 98.362 0.333 0.417 76.467 0.411 0.077 High-End Segment 2002 2003 2004 2005 2006 2007 2008 2009 Q1 68.007 98.352 0.177 0.400 0.428 0.186 Mid range Segment 0.455 87.540 73.711 75.170 Mid range Segment 96.160 99.354 0.167 0.178 0.194 81.238 88.323 0.158 88.244 0.212 0.275 75.595 99.240 99.879 70.429 0.498 0.246 99.395 Table 16: C4 for China by Shipments Years Total market Entry Level segment 68.444 Table 14: C4 for China by Vendor Revenue Years Total market Entry Level segment 71.953 99.358 99.931 83.362 0.170 0.416 0.656 99.253 70.351 0.347 0.714 98.812 80.459 0.867 99.801 74.911 95.464 0.803 94.739 High-End Segment 2002 2003 2004 2005 2006 2007 2008 2009 Q1 83.338 0.164 99.177 0.168 0.266 0.066 99.367 76.174 0.202 0.517 0.477 0.963 69.981 99.733 83.606 99.080 98.255 0.417 0.775 70.275 0.

648 89.185 0.213 0.209 0.568 99.218 0.509 93.243 0.315 0.484 97.196 0.246 97.298 0.219 0.732 95.Korea Table 17: HHI for Korea by Vendor Revenue Years Total market Entry Level segment 0.321 0.299 0.365 0.243 Mid range Segment 0.339 0.881 94.262 0.320 97.314 0.288 0.920 99.284 0.212 0.690 100.308 0.118 97.165 70.171 0.223 Mid range Segment 0.326 High-End Segment 2002 2003 2004 2005 2006 2007 2008 2009 Q1 0.387 82.246 0.363 97.797 95.779 78.223 0.833 71.873 79.396 0.884 95.159 0.668 77.021 73.191 0.223 0.879 82.934 100.601 95.906 100.182 0.000 62 .279 99.288 0.441 69.447 64.895 69.447 83.375 0.182 0.330 0.005 80.185 0.000 96.000 Table 19: HHI for Korea by Shipments Years Total market Entry Level segment 0.209 0.615 85.170 0.206 75.514 Table 18: C4 for Korea by Vendor Revenue Years Total market Entry Level segment 77.224 0.907 94.438 88.513 99.210 0.258 0.319 High-End Segment 2002 2003 2004 2005 2006 2007 2008 2009 Q1 0.487 93.388 0.444 0.017 89.661 100.333 0.569 92.735 69.310 80.558 69.203 0.037 81.162 0.263 0.382 0.246 Mid range Segment 99.000 High-End Segment 2002 2003 2004 2005 2006 2007 2008 2009 Q1 89.455 0.318 0.446 91.370 91.407 Table 20: C4 for Korea by Shipments Years Total market Entry Level segment Mid range Segment High-End Segment 2002 2003 2004 2005 2006 2007 2008 2009 Q1 74.855 76.952 77.139 95.344 93.159 0.935 68.257 0.663 94.186 0.554 83.382 79.382 0.104 95.320 0.059 97.257 0.552 99.122 97.881 66.186 0.415 0.244 0.284 0.319 0.263 0.

008 Mid range Segment 0.010 0.034 0.070 0.015 0.023 0.025 0.027 0.027 0.026 0.028 63 .035 0.029 0.011 0.022 0.028 0.020 0.042 0.046 0.022 0.007 0.012 0.010 0.028 0.009 0.033 Variance estimates on the basis of Shipments Years 2002 2003 2004 2005 2006 2007 2008 2009 Q1 Entry Level segment 0.007 0.008 0.022 0.025 0.010 0.036 0.009 0.024 0.046 High-End Segment 0.039 High-End Segment 0.012 0.021 0.008 0.027 0.049 0.022 0.028 0.Annex VI Variance Variance estimates on the basis of Revenue Years 2002 2003 2004 2005 2006 2007 2008 2009 Q1 Entry Level segment 0.027 0.012 0.010 Mid range Segment 0.039 0.020 0.

0 100.6 93.2 82.0 0.5 91.0 0 0 100.2 98.0 92.2 0 0 100 100 100 100.8 54.0 0 0 100.0 0.0 0.7 54.0 100.4 3. x86 Fujitsu/Fujitsu Siemens IA64 x86 Groupe Bull IA64 HCL Insol x86 Hewlett-Packard IA64 RISC x86 IBM RISC x86 Lenovo x86 Other Vendors x86 SGI RISC x86 Stratus Computer x86 Sun Microsystems RISC x86 Unisys x86 Verari Systems x86 Wipro x86 Zenith Computer x86 2002 100 100 0 0 0 100 100 0 0 0 0 0 100 100 100 0 7.0 0.Annex VII Vendor wise percentage distribution of Processor in India A.6 0 0 0 0 100 100 100 100 2009 100 100 100 0 100 100 100 100 0 100 0 0 100 100 100 7.0 32.0 11.0 0.7 87.9 8.0 0 0 0 0 100 100 100 100 2005 100 100 100 100 0 100 100 0 0 0 100 100 100 100 100 7.7 100.5 88.0 16.0 1.0 0 0 100 100 100 100.0 100 100 100.0 0.8 0 0 100 100 0.0 100 100 100.0 3.4 100.3 85.9 100.0 4.0 84.0 100.0 0.0 100.7 95.0 45.8 92.5 0 0 100 100 0.0 0 0 100 100 100 0.0 39.0 77.8 96.7 100.0 41.1 5 87.0 0.4 100.4 58.2 89.0 45. In Entry Level market Vendor wise CPU Acer x86 Apple Computer RISC x86 Dell Inc.3 3.0 100.0 0.0 0.7 0 0 100 100 100 3.0 11.5 100 100 100 100 100 0.0 9.7 27.7 67.3 0 0 0 0 100 100 100 100 64 .3 0 0 0 0 100 100 100 100 2008 100 100 100 50 50 100 100 100 8.8 0 0 100 100 100 100.0 0 0 100.0 61.8 22.3 100 100 0 0 100 100 100 100 2006 100 100 100 100 0 100 100 0 0 0 0 0 100 100 100 3.8 4.0 100 100 100.5 88.2 100 100 100 100 100 100 100 100 2007 100 100 100 100 0 100 100 100 14.0 100 100 100.7 0 0 100 100 100 5.2 0 0 0 0 100 100 100 100 2004 100 100 100 100 0 100 100 0 0 0 0 0 100 100 100 3.2 100.8 2.0 100 100 100.2 90.2 100.0 72.0 0.1 8.3 91.0 0 0 0 0 100 100 100 100 2003 100 100 100 100 0 100 100 0 0 0 0 0 100 100 100 9.3 0 0 100 100 0.

7 100 21. High End Server market Vendor wise CPU Fujitsu/Fujitsu Siemens IA64 Hewlett-Packard IA64 RISC IBM Other RISC SGI IA64 RISC Stratus Computer RISC x86 Sun Microsystems RISC x86 2002 0 0 100 0 100 100 14.3 23.9 28.7 53.7 5.3 6.3 76.0 2008 0 0 0 0 100 83.8 0 0 0 100 0 100 100 100 0 2007 0 0 100 36.7 100 0 100 0 0 0 100 100 0 2003 0 0 100 0 100 100 15.1 42.2 24.4 0 0 0 100 0 100 100 100 0 2006 0 0 100 23.B.6 100 100 0 0 0 0 100 100 0 2009 100 100 0 0 100 93.8 76.3 85.7 100 100 0 0 0 0 100 94.9 100 100 0 0 0 0 100 100 0 C.7 83.4 25 30.2 84.3 100 80 20 0 0 0 100 86.3 100 0 52.1 13.3 100 0 38.1 0 0 0 0 0 0 100 73.1 67.0 19.7 15.7 26.9 100 25 75 0 0 0 100 100 0 100 100 0 2005 0 0 100 32.4 5.3 2007 0 0 100 100 100 70. Mid range server market Vendor wise CPU Fujitsu/Fujitsu Siemens IA64 Groupe Bull IA64 Hewlett-Packard IA64 RISC x86 IBM IA64 RISC x86 SGI IA64 RISC x86 Stratus Computer x86 Sun Microsystems RISC x86 2002 0 0 0 0 100 0 71.6 2006 0 0 100 100 100 64.6 82.3 2009 0 0 100 70 30 100 16.1 43.4 17.9 35.1 42.9 100 50 0 50 100 100 100 81.3 100 100 0 0 0 0 100 100 0 65 .6 100 100 0 0 0 0 100 94.7 0 100 0 82.9 61.8 84.4 84.3 63.5 11.4 63.1 100 0 56 44 100 0 100 0 0 0 100 100 0 2003 0 0 0 0 100 13.6 100 0 46.6 100 23.9 100 17.7 100 0 57.8 54.2 100 15.2 100 100 0 100 100 0 100 100 0 2004 0 0 100 26.9 47.1 73.1 100 0 100 0 100 100 100 100 0 2004 0 0 0 0 100 44.7 13.7 0 0 0 100 0 100 100 100 0 2008 100 100 100 45.2 100 15.3 16.4 42.3 2005 0 0 100 100 100 43.7 0 100 0 57.6 13.

According to them large scale of operation cannot be ported to open source. Internet Banking– Satyam initially have given the broad vision and written application software in Java for this. This is to make sure if anything happen to Satyam SBI will be able to take back their people and run the same application. Internet Banking and Networking. This led the bank to built extensive branch network in India that included many branches in the low income rural areas. It is also true that enterprise version of OSS is considered to be costly. Networking –for data transfer SBI use their own network. Due to the development of technology cost of Switch came down from 14 lakh/ machine in the year 1987 to 4 lakh/machine in the year 2009. These switches are basically non stop server. but it was losing market share to private-sector banks that had implemented more modern centralised core processing systems. Solution SBI have different IT set up for different types of work. in 2002. Pricing system follows the usual government guidelines128. SBI do not want to be dependent on a single service provider to reduce the possibility of lock in. Over the time depending on the availability of new advanced ATM machine at cheaper cost SBI transship its old ATM machine to the rural branches where number of transaction is less. To remain competitive with its private sector counterparts. After verifying the bid SBI selected HP and TCS to provide the core banking solution. Additionally.Annex VIII SBI Case Study SBI Business Challenges The State Bank of India is the oldest and largest bank in India. with more than $250 billion (USD) in assets. Today SBI has 4 HP superdome hardware and HPUX as OS. Core banking -Recognizing the need for large scale centralized systems expertise in the year 2002. It is the second-largest bank in the world in number of branches. Cost mainly varies depending on the number of dispenser it has. Use of Open Source SBI did not find open source as appealing due to the lack of good service provider especially for enterprise version. 128 hardware and software for 3 years and AMC for 2 years .600 branches and nearly 10. For core banking solution tender was called for hardware and software bundled together where only technical score was considered. SBI began the largest implementation of a centralized core system ever undertaken in the banking industry. ATM – SBI has 8500 TM machines which are knows as switch. SBI has huge subscriber and cannot take any risk of failure only for the sake of cost. A separate tender was called for this and HP ( it was Compaq on that time) got the tender. ATM facility.total 5 years 66 . SBI and its associated banks had computerized their branches in the 1090s. SBI sent some people to be trained in the particular programming language and software application at Satyam. Mainly they have Core Banking. resulting in the availability of banking services at more than 14. a tender was called for core banking solution and Infosys was the only company asked to help SBI. After considering the technical performance ( where both bidders were found eligible for the bid) pricing was taken into account.000 ATMs. Unlike private sector banks. SBI has largest single HPUX installed till date with 120 CPU each. SBI has controlling or complete interest in a number of affiliate banks. SBI has a dual role of earning a profit and expanding banking services to the population throughout India. In fact SBI mention that IBM did not cooperate to perform the scalability test of FNS’s BaNCS on their machine. SBI favored real time processing architecture of FNS’s BaNCS system over that of the IBM consortium’s memo post/batch update architecture. After Infosys backed off TCS & HP and IBM & Altec were the bidders for the tender.

Benefits HP could break the myth that for large number of transaction IBM is the only solution by establishing network of 400 regional processing centers for all metro and urban branches. HP could prove that efficient perfromane at high end computing is not IBM’s monopoly. Cordination among the vendors can pose serious challenge to the big player like IBM with their performance. HP is trying to provide end to end solution to its client by collaborating with EDS but this collaboration is yet to fire. Corodination among the different vendors can incraese the level of competition and bring down the price further. Time Line of State Bank of India and Affiliate Banks’ Core Systems Modernization (2000-09) Why it matters SBI is aware of latest technology available in the market and wanted to deploy the best technology that suitable for their work. Although SBI believe that exploitation is a matter of perception but as a user SBI makes every effort to reduce the possibility of lock in and being exploited by the vendors. Cost Always there is a possibility that one vendor can blame other for inefficient perforomance of the system. 67 . By deploying HP Server SBI bank could reversed the trend of customer attrition and is now gaining new market share. Due to the competition in the market and advancment of technology prices are falling down and they always get a better price than the previous purchase.

0 50.5 26.3 2003 100 17.7 0.0 68 .0 0 0.0 20.2 10.7 0.0 0.0 25.0 50.0 100 50.0 100.0 50.0 2007 100 36.0 48.1 63.3 0.7 Windows Groupe Bull Linux Windows HCL Insol Linux Windows Hewlett-Packard Linux Others 0.0 100 18.0 29.0 100 40.0 78.8 0.6 100 3.0 21.0 70.0 0 0.0 100.0 100.0 50.0 0.2 0.0 100 42.2 100 100.0 0.6 100 4.0 0.6 100 7.0 38.0 0 0.2 2006 100 25.0 0.0 100 41.0 0.0 50.0 0 0.0 68.0 43.0 0.0 100 25.9 0.0 0 0.0 0.7 1.9 41.0 100 50.0 0.0 100 0.7 6.1 Windows Fujitsu/Fujitsu Siemens Linux 100.3 16.9 100 0.1 0.8 0.3 0 0.9 65.0 100 25.8 0 0.7 45.0 0.0 50.0 100 22.0 100 50.0 0.8 0 0.0 100 0.0 100 22.0 0.0 0.0 50.0 100 100.7 0 0.0 100 0.2 25.6 14.2 0.0 0.0 100 100.Annex IX Vendor wise percentage distribution of OS installed at Entry level 2002 Acer Linux 100 6.4 0.0 40.3 0.0 UNIX Windows IBM AIX 20.0 0 0.5 0.0 100 2.0 UNIX Windows SGI Linux 0.6 25.0 50.0 10.0 0.0 0 0.0 50.0 0 0.0 0.0 34.1 41.6 100 11.0 0.0 45.6 0.0 95.9 100 16.8 2008 100 50.0 100 50.0 100 28.0 0.0 100 25.6 100 7.0 27.0 0.0 66.3 0 0.0 50.0 35.9 0.0 0.0 0.0 0 0.0 100 0.1 25.1 100 100.0 70.0 100 30.1 51.5 25.0 100 50.0 50.0 33.0 0.0 50.1 0 0.2 52.0 60.0 75.0 85.8 2004 100 31.0 0.0 100 0.0 0.1 100 3.0 50.0 26.0 0.9 100 50.0 48.1 0 0.0 82.5 0.0 25.0 25.5 0 0.0 100 39.0 100.0 0.5 26.0 50.0 50.0 0.4 83.0 0 0.0 100 14.2 25.0 30.0 0.0 100 50.6 0.3 33.7 Linux Others UNIX Windows Lenovo Linux Windows Other Vendors Linux Others 35.5 0 0.0 100 0.0 0.0 0.8 19.4 100 0.8 100 1.0 25.0 50.0 0.0 34.0 100 0.0 50.0 0 0.0 0.8 40.5 0.0 60.0 100 50.0 100 31.0 100 100.0 28.1 19.1 24.3 0 0.9 48.7 4.0 36.0 100 44.0 100 0.0 UNIX Windows Apple Computer UNIX Dell Inc.0 100 60.0 100.9 0.0 0.0 0.0 2009 100 50.0 83.3 0.0 0.0 100 9.0 100 0.8 100 100.3 0 0.9 2005 100 21.0 100 50.0 100 40.0 0.1 46.8 0.0 0 0.0 63.0 UNIX 100.2 100 100.0 50.0 0.8 5.0 100 50.0 100 25. Linux UNIX 10.0 100 50.

9 23.0 21.5 100 20.0 100 12.0 100 50.7 0.0 14.0 0.3 66.2 33.2 3.7 94.5 71.0 100 21.7 72.0 100 100.3 66.0 0 0.0 10.0 0.3 66.0 Vendor wise Percentage distribution of OS installed in the Mid Range server market 2002 Fujitsu/Fujitsu Siemens Windows Groupe Bull Linux 0 0 0 0 2003 0 0 0 0 2004 0 0 0 0 2005 0 0 100 50 2006 0 0 100 57.9 39.5 87.7 40.3 4.0 81.0 0 0.2 0.5 100 22.0 0 0.1 0.0 100 40.7 100 80 28.0 20 0 0 100 13.0 100 0.0 50.0 UNIX Windows Unisys Linux Verari Systems Linux Windows Wipro Linux Windows Zenith Computer Linux Windows 100.0 50.3 0 100 40.0 28.6 0.0 0.0 100 100.6 100 35.0 0.1 2.9 0.0 100 50.0 0.7 100 47.1 50.0 100 20.8 0 0.8 35.7 16.0 UNIX 100 100 86.0 100 50.0 12.3 0 0 0 100 5.6 0 0.8 58.0 50.6 Linux UNIX Windows SGI Linux 16.0 0 0.0 100 100.5 62.5 42.0 0 0.3 6.7 2.7 0 0 0.0 0 0.3 100 46.0 100 50.5 79.0 100 13.0 50.1 2007 0 0 100 50 2008 0 0 0 0 2009 100 100 0 0 Windows Hewlett-Packard Linux 0 100 3.0 100 18.0 100 33.0 26.6 14.0 80.0 100 50.3 0.6 0 0 0 100 10.0 100 50.8 0.7 74.5 5.0 Others UNIX Windows IBM AIX 12.0 0 0.0 50.0 0 0 0 100 0.5 0 100 100 100 0.0 100 26.0 100 33.0 0.9 12.0 71.0 0 0.2 0.0 100 50.0 0 0.7 24.9 0 0.8 30.0 0 0.1 0 100 6.0 100 27.4 0.0 100 25.0 60.7 44.3 50 100 23.0 100 100.0 86.3 5.0 50.6 100 100 16.0 50.7 70.6 33.0 100 100 100 100 0.0 100 100.7 100 37.2 100 100.3 28.0 100 0 23.3 72.9 100 100 0.0 50.5 100 100 100 69 .0 0 0.3 100 100 11.3 100 28.0 0 0.0 0.0 100 25.9 100 20.4 16.0 100 50.3 100 100 UNIX Stratus Computer Windows Sun Microsystems Linux 100 0 0 100 0.0 0 0.5 100 0 26.0 100 50.4 89.0 0 0 0 100 0.0 53.7 3.0 49.6 100 47.1 0.0 57.7 0 100 33.7 100 28.0 66.0 100 100.0 50.4 0 100 33.5 100 44.0 0.0 54.0 100 33.0 0.3 5.2002 2003 2004 2005 2006 2007 2008 2009 Windows Stratus Computer Windows Sun Microsystems Linux 0.0 50.8 50 100 29.8 4.1 14.7 0.0 0.

0 75.5 18.0 0 100 50 Linux UNIX zOS/OS390 SGI Linux 0 57.Vendor wise percentage distribution of OS installed in the High End Market 2002 Fujitsu/Fujitsu Siemens Windows Hewlett-Packard Linux 0 0 100 0 2003 0 0 100 0 2004 0 0 100 13.0 0 100 48.8 0 100 50 25.0 60.7 100 100 UNIX Stratus Computer Others 100 0 0 0 100 0 0 100 0 0 100 0 0 100 0 0 100 100 0 0 0 0 0 0 UNIX Windows Sun Microsystems UNIX 0 0 100 100 100 0 100 100 100 0 100 100 0 100 100 100 0 100 100 100 0 0 100 100 0 0 100 100 0 0 100 100 70 .0 0 100 28.3 53.2 81.7 23.2 15.0 75.2 0 0 0 26.0 0 100 53.6 21.8 10.8 100 100 0 50 25 0 0 0 52.0 81.7 2006 0 0 100 0 2007 0 0 100 0 2008 100 100 100 0 2009 0 0 100 20 Others UNIX Windows IBM AIX 25.3 100 100 0 33.4 100 29.6 0 0 6.1 30.0 2005 0 0 100 10.8 10 70.6 42.1 30.0 100 25 14.3 16.3 100 0 0 63.1 13.4 19.9 17.9 57.1 14.3 0 0 5.9 13.1 0 100 21.3 15.

Annex X Questionnaire for Server Vendor Company – Respondent– Designation –
Section A Benchmark of Your Product 01. In which market segments are your server products targeted the most? a) b) c) High end Mid range Entry level

02. What is the benchmark you use to define the server market segments (high end, mid range and entry level)? Please provide details. a) b) c) d) Technology (no of processors/cores etc) Price No. of Employees Types of workload

03. Do you have full gamut of products applicable for your target segment/s? a) b) c) High end Mid range Entry level YES YES YES NO NO NO

04. In which regions of India are you active? a) b) c) d) e) f) g) Across India Western Region Northern Region Eastern Region Southern Region Central India Selected cities

05. In which verticals are you active? a) b) c) d) e) f) Banking Insurance Telecom Retail Healthcare Education g) Manufacturing h) Transportation i) Media & Entertainment j) Utilities k) Public Sector/Government l) Any other, Please specify

71

Section B: Factors you consider most while offering your product 06. Please rank the following factors in terms of importance for the Indian Market. Factors
1. 2. 3. 4. 5. 6. Availability (business continuity) Reliability Features & Functions Energy efficiency, (low energy consumption) Cost reductions, (low maintenance cost) Others

RANK on the basis of priority [1-6]

07. What are the key factors (USP) that differentiate your products from the competitors? Please illustrate. Factors Better price-performance Lower TCO Better/Quick ROI Unique features and functions Cost-effective Better Service Interoperability/ Openness Others (please specify) High-end Mid-range Entry-level

08. How frequent are your product refresh cycles? a. b. c. d. e. Less than 3 months 3 months to 1 year 1 – 2 years More than 2 years Depends on sales target achieved

09. What do you give importance to in your product refresh cycles? Please rate on scale of 1-5 [1= Most important, 5=Least important] Factors Technology demand Price competitiveness Inventory Availability Market share leadership Market vision Others Rate on scale of 1-5

72

Section C: Pricing & Bundling 10. A. Do you bundle software with your server? If so what kind of software do you bundle? a) b) c) d) e) f) g) No bundling System software Management software Applications Security software Middleware Others (Provide details)

10. B. If the answer to Q10 A is yes then provide the charging mechanism. a) b) c) d) Unlimited License License per seat Periodic subscription Usage based [SaaS model]

11. When selling your server products do you provide the following options? a) Leasing b) Buy back option

12. Do you sell refurbished server products in Indian market? YES NO Plan to sale in future

13. Do you promote open source? If yes how?

If not, why not?

14. What percentage of your server product lines enables open source software? Server segments High End servers Midrange servers Entry Level servers %

15. In your product pricing, what percentage of price is associated with hardware, software and/or services?

Types of server High end Mid range Entry level

Hardware

Software

Services

Total 100 100 100

73

Do you think there are any unfair trading practices that are hindering your product sales? Please describe.Section D: Competition in the server market 16. 10 -low 19. To do business in India To create level-playing field To reduce product price for end-users Yes Yes Yes No No No B. Do you think there should be restrictiveness on branding. Market Segment High End Middle Range Entry level 17. Do you collaborate with your competitor/s? If so with whom and why? Rate on scale of 1-10. 74 . What level of competition do you face in following segments? Rate the competition level on scale of 1-10. Do you think MNCs have any undue advantage in India? Why? D. Where do you find intensity of the competition more -in hardware.. advertising and other promotional spent to create level playing field? E.1-high. 10 -low Rate on scale of 1-10. A. Do you think providing grants & subsidiaries for setting up operations will help you? If so. Does the government policy help you……. in software or maintenance? Market Segment Hardware Software Maintenance (including services) 18. 1-high. how? C.

In case of rank. CRM etc. Section A: Organization setup and Server environment 1. How many servers does your organization have? a) b) c) d) e) e) 1-9 10-49 50-99 100-499 500-999 1000 or more 3. a) b) c) d) e) f) g) h) IBM HP Sun Microsystems Dell HCL Wipro Fujitsu Other (please specify) 5. telecom etc. What are the main uses of your Server infrastructure? Please tick all that apply. a) b) c) d) e) f) g) h) i) Web hosting Mail server File & Print Server Security Applications IT & Network Management Cross-industry Applications (ERP.) Messaging & Collaboration Others (Please specify) 4. rank all options in order of relevant information in ascending order. Please check the relevant box/es Yes No b. Describe your organization by filling the boxes with relevant information: No. Which vendor/s provide servers to your organization? Please tick all that apply. of Employees No.Questionnaire for Users Instructions: a. Please click where more than options are possible. of Computers Organization sector 2.) Core Applications (Banking. Which server architecture does your organization predominantly use? a) b) c) d) X86 RISC EPIC/Itanium Other 75 . c.

Server Vendor Vendor A Vendor B Vendor C 7. Please Specify Windows NT Unix Open HP /IBM Windows 2000 Red Hat Linux Open Enterprise Windows 2003 Sun Microsystems Solaris Open Enterprise Windows 2008 SUSE Linux Open Enterprise z/OS. Please provide details of your server infrastructure. of Cores No.6. TPF Other (Please specify) 9. of processor 8. Which Database software/s your organization is using? a) b) c) d) e) f) Oracle MySQL Microsoft SQL IBM DB2 PostgreSQL Other (please specify) 10.e. For how many years has your company been using the mentioned servers? Has there ever been change in the vendor. Do you use open source software? If yes identify two major constraints with respect to each Open Source Proprietary 76 . If so when and why? Model/ Family name Number of Servers Clock Speed No. z/VSE. z/OS. Approximately what percentage of your server infrastructure is used to run mission-critical data? a) b) c) d) 1 – 25 26 – 50 51 – 75 76 – 100 11. What Operating System do you run on your servers (Open source or proprietary).

How frequently do you upgrade/replace the existing server infrastructure? a) b) c) d) Less than 3 years 3 – 5 years More than 5 years As and when need arises 77 . who are the consultants? 14. Factors Product features and functions (speed. Purchase decision of server/software is made: a. please rank following factors based on there importance in decision making. etc) Vendor’s reputation & Brand Vendor’s size and financial stability Pre-sales & Post-sales services Total Cost of Ownership associated with the product (CAPEX + OPEX) Expected Return on Investment (ROI) Security Other (please specify) 16. Are decisions made solely on the basis of lowest bid (L1)? If not. scalability. for highest 1 and for lowest 10) a) d) g) j) Integration Market share Previous experience Others b) License price e) Documentation h) Interoperability c) Installer f) Clients support i) Compatibility 18.Section B: Purchasing Decisions & Maintenance Policies 12. Completely in-house by resident experts b. Consultants involved / hired to evaluate tenders If yes. While choosing Operating System/software what criteria are important? (Rank in order of importance. Is security of the server is provided by a) b) The vendor The third Party Rank(1-7) 17. How many quotations are usually invited? Yes Yes No No 15. Do you have your own IT department or do you outsource services? a) b) Own Outsource (mention Outsource service provider/s) 13.

g. d. b. Is your cost of maintenance higher than the cost of buying the bundle? Yes No Not Aware Section C: Customer Satisfaction 22. g. 6)? From Vendor A From Vendor B From Vendor C YES YES YES NO NO NO A. for highest 1 and for lowest 8) Reasons It’s an incredibly secure environment The performance levels are excellent The system is never down Disaster recovery and emergency management are extremely efficient Centralized management is relatively easy The cost per transaction is lower than other systems Low downtime cost Others (please specify) Vendor A Vendor B Vendor C 78 . h. e. Are you satisfied with the performance of your server (mentioned in Que.19. How do you plan to increase utilization of current server infrastructure? a) b) c) d) Load Balancers Virtualization Consolidation Other (please specify) 21. If YES. RISC to x86 or Unix to windows) Others (Please specify) 20. c. Please provide reasons (Rank in order of importance. Financial situation improved in the organization Processor(s) too slow/RAM Capacity too low Company policy on hardware retirement/replacement cycle Other hardware limitations Poor reliability/ high failure rate To save floor space/reduce footprint Migration of server platform (e. f. a. Please state the reasons for upgrading/replacing the existing infrastructure.

Less efficient then what was expected (running cost) AMC.B. Do you think cost of buying a mainframe is too high or. If NO. for highest 1 and for lowest6) Reasons High running cost ( including cooling cost. why is the actual cost higher? Reasons Hidden cost. is the mainframe overpriced? YES NO Not Aware 24. Cost of migration. salary for skilled personal. Did you find that actual cost of using a mainframe is higher than the estimated (or expected) cost? YES NO Not Aware –If your answer is yes. Services others (describe) RANK 79 . please explain why you are not satisfied with the performance of your server? (Rank in order of importance. License. transaction cost ) Not compatible with other application software provided by third party Maintenance cost is high Lack of skilled personal who can run server Services offered by the company is not up to the mark Others (please specify) Vendor A Vendor B Vendor C Applicable for Mainframe User 23.

HP Windows 2000. Crossindustry Applications . Sun Solaris Enterprise version Windows 2000.Annex XI Systems and OS deployment of firms profiled in our sample User Vertical Size of Organisati on 3000 Vendor Use of server infrastructure Operating system Bombay Hospitals Healthcare Max Healthcare Healthcare 4000 Manipal Health Systems Healthcare 7000 IBM. File & printer server. Sun Microsyste ms Microsyste ms IBM. Windows 2003. Sun Microsyste ms Microsyste ms HP. File & printer server. Dell Windows 2003. File & printer server. Windows 2008. Messaging & Collaboration Mail server. Red Hat Linux Wockhardt Ltd Healthcare 7000 IBM. Messaging & Collaboration Mail server. Red Hat Linux 80 . Sun Microsyste ms Microsyste ms HP Windows 2000. Security Applications. IBM AIX. Security Applications. File & printer server. Messaging & Collaboration Mail server. Core Applications. Messaging & Collaboration Web hosting. Messaging & Collaboration Mail server. IT & Network Management. Core Applications. IT & Network Management. Windows 2003 Windows 2003. File & printer server. Core Applications. Mail server. Security Applications. File & printer server. Crossindustry Applications . Cross-industry Applications . Windows 2003. Messaging & Collaboration Windows 2003. Red Hat Linux Hinduja Hospital Healthcare 4000+ Max Healthcare (Delhi) Healthcare 2000 IBM. Red Hat Linux Mail server. Core Applications. Core Applications. Mail server. Sun Microsyste ms Microsyste ms HP Core Applications. Cross-industry Applications . Windows 2008 Shoppers stop Retail 5000 IBM. IBM AIX. Security Applications.Used for Hospital Management software IBM AIX. File & printer server. Core Applications. Core Applications. Sun Solaris Enterprise Fortis Healthcare Healthcare 10000 Windows 2000 Advanced. Windows 2003. HP. Messaging & Collaboration Web hosting. IT & Network Management.

File & printer server. Mail server. IT & Network Management. Messaging & Collaboration Web hosting. Windows 2008. Retail Size of Organisati on 22000+ Vendor Use of server infrastructure Operating system IBM. Messaging & Collaboration Mail server. Red Hat Linux. IBM AIX Retail 1500+ IBM. Cross-industry Applications . Sun Solaris. Windows 2003. Unisys Mainframe Windows 2003. Core Applications Mail server. . Core Applications. Core Applications. Windows 2003. Windows 2003. Messaging & Collaboration Windows 2000. Sun Solaris Windows 2003. HP UX. IT & Network Management. Core Applications Web hosting. HP. Core Applications. IT & Network Management. Sun Solaris Spencers Retail Retail 16000+ Sun Microsyste ms Microsyste ms Windows 2003. (Tata Retail chain) Retail 1600+ IBM Windows 2003. Cross-industry Applications . Core Applications. File & printer server. Messaging & Collaboration Mail server. File & printer server. File & printer server. Messaging & Collaboration Mail server. Red Hat Linux Concor Transporation 1150+ Credence Logistics Transporation 300 Sun Microsyste ms Microsyste ms IBM Windows 2000 Advanced. Mail server. File & printer server. OS 2200 (Unisys Mainframe OS) 81 . Cross-industry Applications . Sun Solaris 10 Croma (Infiniti Retail) Trent Ltd. Wipro Windows 2000. Crossindustry Applications . Red Hat LinuxEnterprise MobileStore Retail 4500+ Aditya Birla Retail Ltd. Crossindustry Applications . HP UX. File & printer server. Retail 11000 HP. HP. Core Applications Mail server. Sun Solaris. IT & Network Management. Sun Microsyste ms Microsyste ms HP. Sun Microsyste ms Microsyste ms Windows 2003. File & printer server. Mail server. HP. Crossindustry Applications . Messaging & Collaboration Mail server. Cross-industry Applications. Red Hat Linux Windows 2003 Air India Transporation 31000 IBM. zO/S. Crossindustry Applications. Sun Microsyste ms Microsyste ms Web hosting. Core Applications.User Vertical Pantaloon Retail India Ltd. IBM AIX. File & printer server. IT & Network Management.

Sun Microsyste ms Microsyste ms Windows 2000. Windows 2003. Security Applications. Core Applications.1. IT & Network Management. HP. IBM AIX 82 . IT & Network Management. Mail server. Red Hat LinuxEnterprise version Central Bank of India Banking 32000 IBM. HP. IBM AIX. Sun Microsyste ms Microsyste ms. Red Hat Linux.User Vertical ICICI Bank Banking Size of Organisati on 30000+ Vendor Use of server infrastructure Operating system IBM. z/OS HDFC Bank Banking 37000+ IBM. IT & Network Management. Core Applications Allahabad Bank Banking 20500 IBM. Cross-industry Applications . Red Hat LinuxOpen source and Enterprise version Windows 2000. IBM AIX. Sun Microsyste ms Microsyste ms Windows 2000. Core Applications. Linux AS 2. File & printer server. Core Applications. Sun Microsyste ms Microsyste ms. HP. Solaris Enterprise version SIDBI Banking 900 IBM. Messaging & Collaboration Mail server. HCL Windows 2000. Cross-industry Applications . Sun Solaris. File & printer server. Acer Windows NT. Messaging & Collaboration Windows 2000. Dell Web hosting. IT & Network Management. HP UX. Messaging & Collaboration Web hosting. Security Applications. HP. Core Applications. HP. HP UX. Security Applications. Novel software. IBM AIX. Mail server. Messaging & Collaboration Web hosting. File & printer server. HP. Sun SolarisEnterprise version. Mail server. Windows 2008. Core Applications. Cross-industry Applications . Windows 2003. Dell Web hosting. HCL. IBM AIX. Security Applications. Red Hat LinuxEnterprise version IDBI Bank Banking 10000 IBM. HP UX. Windows 2003. Cross-industry Applications . Red Hat Linux. IT & Network Management. File & printer server. Mail server. File & printer server. Cross-industry Applications . Messaging & Collaboration Web hosting. HP UX. Windows 2003. Security Applications. Security Applications. Windows 2003. File & printer server. Windows 2000. Mail server.

Messaging & Collaboration Web hosting. Mail server. Mail server. Sun Solaris. Sun Solaris 83 . Messaging & Collaboration Web hosting. Messaging & Collaboration Web hosting. Red Hat Linux. Messaging & Collaboration Windows 2003. Wipro Windows 2000. HP Windows 2000. IT & Network Management. Security Applications. Mail server. Security Applications. Red Hat Linux Reliance Communicati ons Telecommunicati ons 20000+ HP. Security Applications. Sun Microsyste ms Microsyste ms Windows 2003. IT & Network Management. Telecommunicati ons 6100+ IBM. Sun Microsyste ms Microsyste ms. Sun Microsyste ms Microsyste ms Windows 2000.User Vertical BPCL Utilities Size of Organisati on 14000+ Vendor Use of server infrastructure Operating system IBM. Core Applications. Messaging & Collaboration File & printer server. File & printer server. Sun Microsyste ms Microsyste ms Windows 2003. Windows 2003. IBM AIX. Cross-industry Applications . Core Applications. Security Applications. File & printer server. HP UX. Security Applications. File & printer server. File & printer server. HP UX. Windows 2003. IBM AIX. Cross-industry Applications . Mail server. Red Hat Linux. Red Hat Linux HPCL Utilities 14000 IBM. Mail server. Dell Web hosting. IBM AIX. Cross-industry Applications . IT & Network Management. HP. IT & Network Management. IT & Network Management. Red Hat Linux MTNL Telecommunicati ons 47400 IBM. Messaging & Collaboration Web hosting. IBM AIX. Cross-industry Applications . Sun Microsyste ms Microsyste ms. Novell Suse Linux Mahanagar Gas Utilities 300+ HP. Core Applications. File & printer server. Core Applications. HP UX. HP. Cross-industry Applications . True64 UNIX Idea Cellular Ltd. Core Applications. HP UX. Sun Solaris. Crossindustry Applications . HP.

HP UX. Process Manufacturing 2600+ IBM. Suse Linux Enterprise Windows NT. HP UX. Mail server. Messaging & Collaboration Mail server. IBM AIX. HP Asian Paints Process Manufacturing 4700 Windows 2003. Red Hat Linux Enterprise. Crossindustry Applications. Sun Solaris. HP VIP Industries Ltd. HP UX.User Vertical Tata Teleservices Telecommunicati ons Size of Organisati on 5000+ Vendor Use of server infrastructure Operating system HP. HP UX. File & printer server. IT & Network Management. Suse Linux 7. Mail server. IT & Network Management. Crossindustry Applications. Messaging & Collaboration Web hosting. File & printer server. Windows 2008. Cross-industry Applications. Messaging & Collaboration Web hosting. File & printer server. Security Applications. Mail server. File & printer server. Core Applications. FreeBSD Windows 2003. Windows 2003. Core Applications. Messaging & Collaboration Mail server. Windows 2003. Messaging & Collaboration Windows 2000. Windows 2003. HCL Ambuja Cement Process Manufacturing 3500+ HP. File & printer server. Wipro IBM. Security Applications. IT & Network Management. Cross-industry Applications . Sun SolarisEnterprise. IBM AIX Britannia Industries Ltd. IT & Network Management. Red Hat Linux Major Telecom Operator Telecommunicati ons 18000+ IBM. Sun Microsyste ms Microsyste ms Web hosting. Sun Microsyste ms Microsyste ms. HP UX. Messaging & Collaboration Mail server. Sun Microsyste ms Microsyste ms Windows 2000. HP. IBM AIX Vodafone Telecommunicati ons NA IBM. File & printer server. Core Applications Mail server. HP.0. Marico Industries Process Manufacturing Process Manufacturing 1600 IBM Windows 2003. File & printer server. File & printer server. Crossindustry Applications . Sun Microsyste ms Microsyste ms Windows 2003. Core Applications. Red Hat Linux Enterprise 84 . HP UX. Crossindustry Applications Web hosting. IBM AIX Windows 2003. Red Hat Linux Enterprise. HP UX 1190 HP. IT & Network Management. Mail server. IT & Network Management. Cross-industry Applications . Cross-industry Applications.

Core Applications Windows 2000. HP. Mail server. HP UX. Core Applications Web hosting. IT & Network Management. Core Applications Web hosting. HCL. Sun Microsyste ms Microsyste ms Oriental Insurance Company Limited General Insurance Corporation of India Aviva Life Insurance Insurance 16000 Insurance 450+ Sun Microsyste ms Microsyste ms IBM. IT & Network Management. Windows 2003. IBM AIX. File & printer server. IT & Network Management. Security Applications. Mail server. Mail server. Mail server.open source. Red Hat Linux Enterprise Windows 2000. Security Applications. Cross-industry Applications . Cross-industry Applications . File & printer server. Core Applications. Mail server. Windows 2003.Open. Mail server. HP UX.Core Applications. Messaging & Collaboration Web hosting. Windows 2000. Security Applications. Red Hat Linux 85 . HCL Windows 2003. HP. Core Applications. Red Hat Linux MSRTC Government 1. Core Applications Windows NT. IBM AIX. Security Applications. File & printer server. HP. HP. Sun Solaris. File & printer server. Dell Windows 2000. Dell LIC Insurance 70000 IBM. Unix . Windows 2008. Windows 2003 Advanced Version. HP UX. Red Hat Linux.User Vertical IFFCO Process Manufacturing Size of Organisati on 6728 Vendor Use of server infrastructure Operating system IBM. HP. Security Applications. Windows 2003 Maharashtra State Electicity Board Government 75000 IBM. Acer Web hosting.1 Lakh IBM. Sun Microsyste ms Microsyste ms Web hosting. Crossindustry Applications . Messaging & Collaboration Bajaj Allianz Insurance 32000+ IBM. Red Hat Linux. Core Applications Web hosting. HP UX. Messaging & Collaboration Web hosting. Linuxopen source Windows 2000. IT & Network Management. Cross-industry Applications. Cross-industry Applications . z/OS and IFL Sun Solaris Enterprise verison Insurance 2500+ IBM Web hosting. Crossindustry Applications . File & printer server. Mail server. Unix . Windows 2003. Sun Solaris. Novel Suse Linux Windows 2000. Mail server.

IBM AIX. Wipro IBM. Security Applications. HP. IBM AIX. Sun SolarisEnterprise version. HP CRIS Government 500+ IBM. Red Hat LinuxEnterprise version Windows NT.Open. IBM AIX. Mail server. Mail server. Security Applications. Red Hat LinuxEnterprise Mausam Bhawan Government 8000 RBI Government >20000 Web hosting. IT & Network Management. Core Applications Windows 2003. Red Hat LinuxEnterprise version Windows NT. Messaging & Collaboration Windows 2003. Security Applications. OthersERM/CITRIX/CODEX/ DNS Web hosting. HP UX. Mail server. Dell. HP. Messaging & Collaboration Web hosting. File & printer server. Security Applications. Windows 2000. Sun Microsyste ms Microsyste ms. Windows 2008.User Vertical CDAC Government Size of Organisati on 800 Vendor Use of server infrastructure Operating system IBM. Windows 2003. Red Hat Linux. HCL. IBM AIX. IT & Network Management. Red Hat LinuxEnterprise version. HCL IBM. HP. OthersR&D Department of Trade and Taxes Government 60200 IBM. Sun Microsyste ms Microsyste ms Web hosting. HCL. Mail server. Windows 2003. Dell. z/OS 86 . Sun Microsyste ms Microsyste ms. Core Applications. File & printer server. IT & Network Management. HP.Open source and Enterprise version. . Sun SolarisEnterprise version. . IT & Network Management. Sun Microsyste ms Microsyste ms. Unix . Messaging & Collaboration. Wipro Web hosting. Sun SolarisEnterprise version. Windows 2008. File & printer server. Windows 2000. Dell. Core Applications. SUSE Linux open Windows 2003. HP UX.

NEC Calculating & Accounting Machines (No Electronic Computers) Electronic & Other Electrical Equipment (No Computer Equip) Wholesale-Computers & Peripheral Equipment & Software Retail-Computer & Computer Software Stores Services-Computer Programming. The delicensed undertakings.g. There are no restrictions regarding the location of the manufacturing unit. Services-Computer Programming Services Services-Prepackaged Software Services-Computer Integrated Systems Design Services-Computer Processing & Data Preparation Services-Computer Rental & Leasing NIC CODE: Division 72 Computer and Related Activities Trade Policy Electronic goods and computer hardware and software are classified under HS code 8505 and 8515 to 8545. Etc. There are few items which require import licences. Industry Policy Industrial undertakings manufacturing electronic goods and computer hardware are exempt from obtaining an industrial license to manufacture. clock and timing circuits. Custom Duty Full exemption from custom duty on computer software Tariff Item 8541 50 00 8541 60 00 8541 90 00 8542 8542 31 00 Description of Goods Other Semi-conductors devices Mounted Piezo-electric crystals Parts Electronic Integrated Circuits Processors and controllers. Rate of Duty Free Free Free Free 8542 32 00 8542 33 00 8542 39 00 8542 90 00 U U U K. amplifiers. whether or not combined with memories. are required to file an Industrial Entrepreneur Memoranda (IEM) with the Secretariat of Industrial Assistance (SIA). Free Free 7.50% Free .g. No further approval is required. Imports of most items are free.Annex XII SIC Code 3570 3571 3572 3575 3576 3577 3578 3600 5045 5734 7370 7371 7372 7373 7374 7377 Industry Title Computer & office Equipment Electronic Computers Computer Storage Devices Computer Terminals Computer Communications Equipment Computer Peripheral Equipment. or other circuits Memories Amplifiers Other Parts 87 Unit U U K. Data Processing. converters. logic circuits. however.

Cost According to the deal IBM is precluded for two years from selling the technology developed for Bharti Airtel. 129 IDEA is a private telecom operator in India. It is always challenging to maintain a wide service footprint and at the same time provide high level of customer service. it needs reliable and secure IT infrastructure. higher customer satisfaction and more profitable growth. Solution Airtel signed a 10 year. IBM provides all hardware and software and services as per the requirement of Bharti airtel. Although popularity of this revenue sharing model largely depends on the success of the business partner. Although in a revenue sharing model benefits extracted depend on the bargaining power along with the growth of the player. Why it matters Clearly in this business model IBM will always have incentive to run the system smooth and flawless to help Airtel grow business. This deal gave an opportunity to provide end to end solution and set an example in the market.5 billion. both IBM and Airtel see this model as an opportunity to grow faster than their corresponding competitors. although IBM can replicate the business model with other telecom player but they cannot use the same technology before the completion of two years from the date of installation in Airtel. This model is the first of its kind in the world and IBM is trying to replicate the same model with IDEA129. Like all other TELCO at initial stage Airtel used to have Sun Microsystems server with Sun UNIX as OS.Annex XIII Bharti Airtel.Case study Bharti Airtel Business Challenge Bharti Airtel is India’s largest private sector telecom operator and only one to offer its services in each of India’s 23 “circles”. IBM is taking initiatives to help Bharti Airtel grow and increase its own revenue. Benefits This model will provide Bharti Airtel to avoid huge capital expenditures and giving IBM an opportunity to lease its hardware. Use of Open Source IT infrastructure is maintained by IBM and they deployed software mainly written by IBM and its partner vendor which are proprietary software. server vendors started considering this model as an alternative to the existing CAPEX model. Over the years to offer better services to its rapidly growing base of subscriber and to meet the challenge of financial risks from a steady decline in India’s average revenue per user (ARPU) for mobile telecom services Airtel decided to outsource its IT infrastructure hence move from CAPEX model to OPEX model. According to the deal IBM will always provide the latest technology require meeting Bharti Airtel’s growing business need by replacing the old IBM hardware. $750 million revenue-sharing deal in 2004 with IBM to manage its core IT infrastructure. helped by robust growth in Bharti’s subscribers and revenue. The deal is now estimated to have gone up to more than $1. 88 . This deal gave IBM an opportunity to secure a flow of income instead of receiving a lump sum once. Outsourcing of technology enables Bharti Airtel to focus resources on growing business. In order to keep up all the processes required to run its business-from order management and service activation to those processes involved in the operation of its core networks. According to the deal. Its new platform provides a standardized framework for Bharti Airtel to integrate its channels and customer-facing processes-enabling a more seamless customer experience. Currently Airtel is using more than 3000 servers with on an average 16 processor in each server.

At the time IBM sold its hardware. Edelstein to "conduct no further proceedings of any kind whatsoever with respect to the parties and issues in the case of United States v. By 1955.. For nearly 50 years the company operated under an agreement with the government that sought to limit IBM’s power in certain markets130. the trial of U. That is. Baxter signed a Stipulation of Dismissal that stated the government's charges were “without merit. So in the July 1969 IBM signed another consent decree to unbundled which led to the development of hundreds of companies for supplying software (like University Computing and Computer Associates). specifically computers designed primarily for business.” 89 . International Business Machines Corp. The suit alleged that IBM violated Section 2 of the Sherman Act by monopolizing or attempting to monopolize the general purpose electronic digital computer system market. attempted to monopolize and restrained trade in the tabulating industry. training. a so-called antitrust consent decree. Among other things. the decree encouraged the development of an independent repair and maintenance service industry to support the used equipment market. or experimentation with such machines. “Every contract combination in the form of trust or otherwise or conspiracy in restraint of trade or commerce among the several states or with foreign nations is declared to be illegal. alterations in. For example. if someone wanted the mainframe software they also had to purchase hardware. It also covered computers. 131 Section 1 of the Sherman Act says. 1982. In January 1969 the US Justice Department brought an antitrust action against IBM for monopolizing the computer market.Annex XIV Examples of Mainframe legal cases IBM does dominate when it comes to supplying mainframe systems that run on its own operating system by not licensing its z/OS on non-IBM hardware.. IBM began and spanned a period of over six years. 1975. Department of Justice.” 130 IBM Consent Decree: The primary purpose of the decree was "to establish in the United States a used machine market" to compete against IBM's new computers. and required grant backs of any inventions resulting from a breach of the prohibition on experimentation. Through its lease agreements. to preserve documents. in violation of Sections 1 and 2 of the Sherman Act131. The decree requires IBM to sell its computers as well as lease them and to service and sell parts for computers that IBM no longer owned. and parts of it gradually phased out until all provisions were dropped in 2001. On January 21. the Government alleged that IBM had monopolized. limited machine uses. and everything else from IBM. the Complaint alleged that IBM only leased. was struck in 1956 to settle allegations of monopoly abuse in the market for electronic tabulating machines.” Section 2 of the Sherman Act says that “Every person who shall monopolize or attempt to monopolize or combine or conspire with any other person or persons to monopolize any part of the trade or com. on January 8.. v. An unanticipated consequence of the decree was the development of computer lessors that competed with IBM in financing. and the like).” and to vacate certain orders he has entered “requiring the parties . restricted attachments to. After 13 years of disputes. hardware (disk drives. The agreement. IBM allegedly: charged lessees a single price for machine rental and repair and maintenance. training. memory. and refused to sell. and all services as a bundled product. On May 19. IBM had adopted the same lease-only strategy with respect to computers.S. IBM petitions for a writ of mandamus directing Judge David N. software. 1952. assistant attorney general in charge of the Antitrust Division.merce among the several states or with foreign nations shall be guilty of a felony.” On August 13. Baxter. tabulating machines. the case was withdrawn by William F. 1982. The other decree provisions generally were intended to reinforce the sale requirement.

Just before the deal was to close. tortuous interference with its economic expectancy and various state law violations. In November 2007. But as the legal proceedings dragged on.” 90 . IBM has since beginning aggressively sought to protect its hold over the mainframe by buying or striking deals with small competitors who raise antitrust concerns.” H.P. among other things. The suit sought injunctive relief and unspecified monetary damages. protecting its cash cow at the expense of consumers. The Department made it clear that if IBM engaged in further “anticompetitive activity” they would be “immediately liable to suit.B.’s lawyers discovered letters from I. including Intel Capital and Goldman Sachs. CCIA described the IBM-Platform deal as “a clear attempt by I. Platform got a jolt of cash when Microsoft joined Platform’s existing investors. 2002. Compuware filed suit in the United States District Court for the Eastern District of Michigan against International Business Machines Corporation ("IBM") alleging. Most of the server rivals were small and dead. bought Platform in July for $150 million. (ii) IBM made false representations regarding Compuware’s software products in violation of the Lanham Act. “We were six to nine months from getting a new product to market. “The investors just didn’t have the stomach for fighting a very difficult case. and (iii) IBM was using its monopoly power to engage in unlawful tying arrangements and was subverting competition on the merits by denying critical information to Compuware and others in an effort to undermine Compuware’s development efforts.B. Compuware claimed that (i) IBM had copied and misappropriated portions of its mainframe software tools and had wrongfully used its technology to develop competing products.P.’s expensive mainframes. 2005. By the late 1990s.M. and in 2006. it fell apart when H. IBM filed patent infringement claims against Compuware in the United States District Court for the Southern District of New York.B. Compuware extracted a $400 million settlement from IBM.M. And then it promptly terminated the innovative product. allowing it to rehire staff and work on a fresh product. On March 22.M. Big Blue fought back.’s custom chips and software. I. to purchase a company solely to foreclose competition in the mainframe marketplace. stating that it would refuse to license its mainframe software to Platform. So when an upstart. antitrust law violations. After legal action failed. recalled. however. Platform Solutions developed software that turned standard servers into systems that mimicked I. In 2004. the former general counsel at Platform. In 2001 the consent decree of IBM was completely dissolved by US Department of Justice. among other things. IBM seeks for the second time an order from this Court directing Judge Edelstein to recuse himself from further proceedings. infringement of its copyrights and misappropriation of its trade secrets with respect to its mainframe software tools. it almost bought the company for close to $200 million.B.Alternatively. liked Platform’s concept.M. all of the other mainframe makers decided to abandon the technology because it was too expensive to keep up with I.” On March 12.M. It is worth noting that most of the case against IBM has gone on for years of battle with different statement of objections but ends with no final resolution. intentional interference with contractual relations with its employees and former employees.B. to put $37 million more into the company.” Gregory Handschuh. Platform’s investors grew weary. Moreover. from IBM. This may be partly because of the complexity of the technical issues. the financial strength of IBM or government’s laissez-faire approach.

In early July 2008 after it was reported that T3 was preparing to file an antitrust complaint of their own in Europe. a Florida-based company bills itself as “The Other Mainframe Provider” is among the firms filing complaints with the government. Mainframe sales are actually surging. While Platform has disappeared. IBM has spent great time and expense developing its technology and will defend its intellectual property rights vigorously. At the same time that it sued Platform. filed a formal antitrust complaint with the European antitrust regulators against IBM. QSGI had no choice but to exit the business of providing refurbished mainframe computers that competed with IBM. still about 25% of co. lives on in a modified form.M. As a result. IBM has not seen T3’s alleged EU complaint. I. IBM acquired PSI who then agreed to terminate its legal proceeding against IBM. European officials announced that they would continue to investigate IBM’s actions in the mainframe marketplace.B. IBM agreed to refer its existing mainframe customers to QSGI for auditing and data erasure services. In January 2009 T3 Technologies. will change its stance.M. Nonetheless. Fundamental sits in limbo with a once-popular product it cannot sell. T3. declined to renew a patent license with Fundamental Software. and if the deal occurred. I.B. More recently. In November 2007.B. its fight against I. Sun Microsystems. I.’s fire. Also. hoping that I.M. has also tried to buy Sun for about $7 billion. The requests for information followed a complaint filed by CCIA. QSGI informed investors that because of IBM’s anti-competitive practices. —a company who refurbished used IBM mainframe computers and sold them to customers for significantly less than the cost of a new IBM mainframe computer—announced during an earnings call that it experienced a sharp decline in refurbished IBM mainframe sales because IBM had terminated its pre-existing policy of providing QSGI with the necessary tools to refurbish and resell IBM mainframes.B.M. which also made mainframe emulation software. while keeping QSGI out of the mainframe market. revenue and nearly half of their profits come from system z mainframe. acc to Wall Street analysts. Hewlett-Packard and Microsoft have made mostly unsuccessful attempts to pull mainframe customers away from I. QSGI.B. by creating products that handle similar tasks but run on servers. The complaint carries the allegation that IBM is blocking competition by offering its operating system along with its mainframe hardware and “abusing its monopoly power in the mainframe industry”. in 2007. the biggest packager of Platform’s technology. T3 Technologies signed onto PSI’s lawsuit against IBM and added its own antitrust allegation. QSGI’s CEO also told investors that its attorneys advised QSGI that IBM’s conduct ran afoul of established antitrust laws.M.B. The Justice Department has started a preliminary antitrust inquiry of IBM by seeking information about IBM’s business practices from companies that compete with IBM in the market for high-end. I. 91 .M.Platform was not the only potential competitor that drew I.M. Under the new partnership. Before QSGI could file an antitrust suit. Despite IBM’s purchase of PSI and the subsequent withdrawal of its antitrust complaint. would also gain a monopoly on the key storage systems used for mainframes.B.B.M. The legal practices against IBM have resulted in higher costs for mainframe. is carrying on the battle with financial support from Microsoft. T3 Technologies. IBM entered into a “partnership” with QSGI for providing mainframe services. IBM is confident that it is no violation of competition laws for IBM to rightfully seek to prevent another company from violating IBM’s intellectual property rights. Mainframe has benefited more from the lack of competition than from updated technology.

IBM doesn’t have many rivals anymore that make mainframe computers.IBM. Most of the legal cases are filed against IBM in the IT server market. They allege that IBM. stopped doing so in recent years to choke off competition. contends that there is no violation of competition laws for IBM to rightfully seek to prevent another company from violating IBM's intellectual property rights. in a statement. but some smaller companies are trying to develop technologies that would allow the software to run on cheaper hardware. which used to license its mainframe software to competitors and for the back half of the last century operated under an antitrust agreement with the government. The accusations stem from claims by IBM rivals that they’ve been illegally frozen out of the mainframe market because of IBM’s refusal to allow its mainframe operating software to run on non-IBM computers. 92 . IBM has spent great time and expense developing its technology and will defend its intellectual property rights vigorously.

3. IBM will no longer support the use of z/OS on Amdahl’s and Hitachi’s IBM-compatible mainframes. In September of 2004. it will discontinue supporting z/OS versions that run on anything other than 64-bit hardware. 93 . and for several years IBM’s OS/390. which combined to account for roughly 21 percent of the mainframe computer market by 1999. 5. 2004.Annex XV IBM’s Evolving Mainframe Operating System 1. IBM announced that. 2. In December 2002. other IBM-compatible mainframe computers supplied by other computer developers such as Amdahl and Hitachi Data Systems. which are 31-bit. IBM upgraded its OS/390 operating system to the z/OS which was compatible with all existing IBM-compatible mainframe software. Accordingly. as of March 2007. 4. leaving z/OS as the only version of the IBM-compatible mainframe operating system in production and serviced or supported by IBM. Initially. the leading operating system marketed by IBM was OS/390. IBM withdrew marketing of the superseded OS/390 version of its operating system and announced that it would discontinue service for OS/390 by September 30. In Jan 2001. From 1996 to 2000.

Annex XVI Forecast: IT expenditures by verticals ($ Millions) Verticals Communications Financial Services Healthcare Government Process Manufacturing Retail Trade Transportation Utilities Total 2006 4828 4362 659 2563 1056 460 656 649 20750 2007 5937 5427 820 3200 1295 582 842 833 25956 2008 6539 6268 939 3703 1475 676 996 977 29820 2009 7104 6821 1031 4074 1584 744 1124 1091 32797 2010 8476 7904 1178 4718 1828 866 1326 1288 38420 2011 9974 8994 1328 5373 2043 990 1543 1501 44175 2012 11371 10139 1484 6010 2266 1135 1774 1741 49949 94 .

8 µ 0.6 µ 0. extended the concept of processor identification and utilized multiple low-power states to conserve power during idle times. and 3D content performance and new power management features will extend “Penryn” processor family leadership Intel's latest-generation micro architecture 8 9 10 2000 2002 2002 1.5 GHz 1.5 µ 1µ 0.com/photogallery/IntelChips1971to2007 95 .000 transistors—more than 100 times as many as the original Intel® 4004. No. This optimizes mobile microarchitecture of the Intel® Pentium® M processor With more than 1. With new SSE4 instructions for improved video. these processors double the performance of previous Itanium processors while reducing average power consumption. imaging.8 GHz 731 million 4 45nm To be announced in late 2009 or early 2010 Source: http://gizmodo. It executes 112 million commands per second This makes possible advanced 3D visualization and interactive capabilities This executed Internet Streaming SIMD Extensions.7 GHz 1 GHz 42000000 5500000 220000000 1 1 2 0. It is theoretically capable of performing roughly eight times more work per clock cycle than other CISC and RISC architectures.7 billion transistors and with two execution cores.66 GHz 291000000 172000000 0 2 2 65 nm 90nm 13 2007 › 3 GHz 820000000 4 45 nm 14 15 15Nov-08 2.000 9500000 No of Cores 1 1 1 1 1 1 1 Manufacturing technology 10 µ 3µ 1.13 µ 11 12 2006 2006 2. The Intel486™ introduced the integrated floating point unit.000 5. This processor ushers in the advent of the nanotechnology age Intel® Core™ processor family delivers advanced mobile performance and is built for energy-efficient power savings The architecture is based on Explicitly Parallel Instruction Computing (EPIC).18 µ 90 nm 0.Annex XVII Intel Chips from 1970 to present S.93 GHz 1.00.18 µ Distinguish Features This is the first general purpose microprocessor that could be customized with software to perform different functions on different devices This was the first 16 bit processor and delivered about ten times the performance of its predecessors The Intel386™ processor could run multiple software programs at once and featured 275.500. 1 2 3 4 5 6 7 Intel-Chips Intel@ 4004 Processor Intel @ 8086 processor Intel@ 386 processor Intel486™ processor Intel® Pentium® processor Intel Pentium Processor Intel® Pentium® III processor Intel® Pentium® 4 processor Intel® Pentium® M processor Intel® Itanium® 2 processor Intel® Core™2 Duo processor Dual-Core Intel® Itanium® 2 processor 9000 series Quad Core Intel® Xeon® Processor (Penryn) Nehalem Mobile – Nehalem Year 1971 1978 1985 1989 1993 1995 1999 Intel clock speed 108 KHz 5 MHz 16 MHz 25 MHz 66 MHz 200 MHz 500 MHz No of transistors 2300 29000 275000 1200000 31.

results in harmful impact for both consumers and other existing or potential competitors. (f) The above instances of ‘abuse’ will need to be tested against the principles of the Competition Act only if it is established through evidence that the enterprise engaging in such abuse is in a ‘dominant position’. as explained above. abuse of dominant position is imperative.Annex XVIII FRAMEWORK FOR INVESTIGATING ABUSE OF DOMINANCE IN INDIA UNDER THE COMPETITION ACT The Act defines: (a) Dominance and (b) Abuse of Dominance. The instances of ‘abuse’ outlined in Box 1 above. BOX 1: ABUSE OF DOMINANT POSITION: Section 4(2) There shall be an abuse of dominant position if an enterprise or group:(a) (b) (c) (d) (e) directly or indirectly imposes unfair or discriminatory conditions or prices in purchase or sale of goods or services. Situations that would amount to an ‘Abuse of dominant position’ have been defined in Section 4(2) of the Competition Act. limits or restricts technical or scientific development relating to goods or services to the prejudice of consumers.’ The Competition Act recognizes that dominance in itself is not a matter of concern. makes conclusion of contracts subject to the acceptance by other parties of supplementary obligations which. by their nature or according to commercial usage. have no connection with the subject of such contracts. Any act of abuse has to be seen in the context of the ‘Relevant Market’. The following Box enumerates these circumstances. Any inquiry into a matter involving an enterprise’s dominant behavior will therefore need to look at each of these concepts as applicable in a given fact situation. These instances can be categorized 96 . (including through “predatory pricing”). the enterprise/group is engaging in abusing its position through conduct as listed in section 4 of the Act. The Act further lists under section 19(4) criteria that are to be considered by the CCI while making an inquiry into whether an enterprise/ group is in a ‘dominant position. limits or restricts production of goods/services or market for such goods or services. or uses its dominant position in one relevant market to enter into or protect other relevant market. indulges in a practice that can deny market access in any manner. The focus of the CCI’s inquiry therefore is the question whether by being in a dominant position in the relevant market. For an act to be in contravention of the law. The framework of inquiry on Abuse of Dominance in any given case will therefore need to comprise of the following elements: (a) (b) Whether the enterprise is in a dominant position in the relevant market? Whether the enterprise is abusing its dominant position in the relevant market? The term ‘dominant position’ is defined under the Act.

Vertical integration of the enterprises or sale or service network of such enterprises.into two broad categories: (a) Exploitative Conduct by an enterprise/ group that results in exploitation of consumers. the CCI would have to decide which aspects are relevant for application. high cost of substitutable goods or service for consumers. It is important to note at the outset that the Competition Act does not mandate that all the factors in section 19(4) need to be applied in all cases. The following box encapsulates the factors listed for consideration by the CCI in an analysis of dominant position. by way of the contribution to the economic development. Relative advantage. have due regard to all or any of the following factors: (a) (b) (c) (d) (e) (f) (g) (h) Market share of the enterprise Size and resources of the enterprise. and depending on the facts and circumstances of each case. no one factor in itself can lead to a complete analysis. financial risk. whereas exclusionary conduct is that which seeks to preserve or expand the power of the enterprise by excluding competitors from entering the relevant market. Entry barriers including barriers such as regulatory barriers. (i) (j) (k) (l) (m) 97 . Economic power of the enterprise including commercial advantages over competitors. Social obligations and social costs. The list of factors in the section provide a guide-post for analysis. high capital cost of entry.19(4): The Commission shall. Any other factor which the CCI may consider relevant for the inquiry. through which the enterprise/ group seeks to anti-competitively impair competition from another enterprise or growth of such other enterprise. Dependence of consumers on the enterprise. Countervailing buying power. economies of scale. At the same time. and (b) Exclusionary conduct by an enterprise/ group. marketing entry barriers. BOX 2: FACTORS FOR ASSESSMENT OF DOMINANT POSITION S. by the enterprise enjoying a dominant position having or likely to have an appreciable adverse effect on competition. Section 19(4) of the Act lists several factors that the CCI would need to consider in determining whether an enterprise or a group is in a dominant position. Monopoly or dominant position whether acquired as a result of any statute or by virtue of being a Government company or a public sector undertaking or otherwise. Size and importance of the competitors. Exploitative conduct is therefore conduct designed to exploit the power of an enterprise. while inquiring whether an enterprise enjoys a dominant position or not under section 4. Market structure and size of market. technical entry barriers.

02 1.03 9.82 0.01 4.13 39.16 14.80 0.47 0.01 8.01 7.00 1.02 2.00 4.18 0.47 0.85 0.96 2.04 0.22 100 2008 1.92 0.98 31.17 0.93 0.45 0.01 15.00 5.08 0.00 0.63 26.00 1.09 0.00 6.81 19.01 1.02 6.60 100 2007 1.13 0.21 20.37 100 2004 3.77 0.16 100 B.12 0.01 0.04 11.44 0.31 0.68 0.20 0.71 0.15 30.01 0.81 1.00 0.25 0.03 7.91 0.68 0.45 0.90 0.53 0.00 0.00 0.24 31.01 8.08 0.81 0.46 0.05 7.01 0.16 0.42 0.00 0.06 5.89 33.25 0.00 3.50 0.78 38.09 2.38 0.02 9.00 2.06 12.00 0.01 3.25 0.37 0.00 1.00 15.06 0.16 0.28 0.31 100 98 . Vendor wise market share for Total Market in India Vendors \ Years Acer Apple Computer Dell Inc Fujitsu/Fujitsu Siemens Groupe Bull HCL Insol HP IBM Lenovo Other Vendors SGI Stratus Computer Sun Microsystems Unisys Verrari Systems Wipro Zenith Computer Total 2002 3.00 3.31 0.67 34.00 0.93 0.04 0.01 11.00 0.27 0.00 12.00 14.00 0.67 31.00 0.38 100 2007 0.00 8.22 100 2009 Q1 0.08 0.98 0.59 0.00 0.38 1.00 1.00 10.19 0.00 0.38 40.00 0.09 0.48 31.11 0.63 0.05 27.00 0.61 0.04 33.25 0.76 0.46 0.00 0.00 5.00 6.74 0.40 0.00 0.99 26.82 100 2004 2.00 1.05 1.97 29.38 29.00 0.00 1.05 35.04 3.75 13.00 0.00 6.74 0.00 9.00 3.10 0.00 0.01 30.37 100 2008 2.49 0.03 11.00 0.00 6.05 8.74 0.00 0.00 2.67 0.07 0.00 13.01 14.21 0.78 0.98 0.00 0.00 10.00 4.57 0.39 0.56 35.13 100 2009 Q1 0.62 0.06 5.00 5.00 0.00 1.01 13.00 0.74 31.33 0.60 31.00 8.02 6.00 4.02 6.00 10.61 33.53 100 2003 7.00 0.94 30.29 24.93 25.00 2.Annex XIX A.00 9.00 0.00 6.24 0.02 100 2006 2.00 2.00 0.00 0.00 0.27 0.00 0.97 0.00 0.92 0.77 100 2003 4.04 0.86 24.78 0.00 2.71 0.37 0.87 0.38 0. Vendor wise market share for Entry Level Segment Vendors \ Years Acer Apple Computer Dell Inc Fujitsu/Fujitsu Siemens Groupe Bull HCL Insol HP IBM Lenovo Other Vendors SGI Stratus Computer Sun Microsystems Unisys Verrari Systems Wipro Zenith Computer Total 2002 6.32 34.01 0.93 31.16 0.25 0.17 0.02 0.37 0.24 24.25 0.43 0.02 23.87 0.00 1.00 0.03 21.13 33.01 0.15 0.00 3.72 0.62 0.25 100 2005 2.99 0.75 0.00 1.18 0.04 0.74 1.01 0.37 100 2005 1.06 0.00 12.78 0.27 0.19 14.19 18.12 0.11 0.00 1.00 2.02 0.29 0.63 100 2006 1.39 0.13 0.71 0.

08 17.71 0.47 18.48 0.04 32.40 28.00 45.37 0.00 0.79 48.83 37.00 29.47 0.21 38.23 34.42 100 2007 0 0.77 0.07 0.47 36.90 100 36.45 0.61 31.89 35.65 100 2005 0.20 100 2009 Q1 1.00 1.00 2008 0.C.91 100 2004 0.00 2.00 7.07 0.39 43.00 16.10 0 42. Vendor wise market share for Mid-range Segment Vendors \ Years Fujitsu/Fujitsu Siemens Groupe Bull Hewlett-Packard IBM SGI Stratus Computer Sun Microsystems Total 2002 0 0 25.25 2009 Q1 0.15 0 18.49 0.45 100 0.21 38.04 0 27.59 100 D.00 0.83 0.79 28.00 35.25 20.01 0 52.72 10.84 48.56 100 2003 0.00 33.00 28.25 1.89 100 2008 0 0 37.57 12.71 100 2006 0 0.91 48.16 35.81 0.30 32.54 0.29 38.08 28.40 100 2007 0.85 40.00 44.26 0.97 100 2006 0.04 1.90 0.39 44.73 33.87 100 99 .93 0.02 0 0 36.29 0 14.58 46.50 0. Vendor wise market share for High-end Segment Vendors \ Years Fujitsu/Fujitsu Siemens Hewlett-Packard IBM SGI Stratus Computer Sun Microsystems Total 2002 0.73 0.71 100 2005 0 0.43 100 2003 0 0 26.00 44.81 100 2004 0 0 33.38 0.52 0 35.

8 72.2 UNIX 4.1 Other 0.1 UNIX 8.5 14.6 Total 100 100 2.9 0.0 Windows 88.0 UNIX 10.9 51.0 100 21.1 UNIX 6.1 58.3 UNIX 8.4 67.2 Windows 77.0 6.2 UNIX 6.7 Windows 77.3 100 14.4 Total 100 100 Education Linux 18.7 Total 100 100 Discrete Manufacturing Linux 9.2 Other 0.5 Other 0.7 14.5 66.9 0.1 11.8 10.6 7.1 Other 0.5 0.9 100 7.006 Process Manufacturing Linux 7.6 83.7 100 10.0 6.0 76.1 11.5 100 20.3 85.3 Windows 82.1 13.6 76.4 Total 100 100 Financial Services Linux 9.0 3.8 Other 0.2 8.6 100 12.1 UNIX 7.4 16.3 Windows 74.4 82.5 4.008 10.6 Total 100 2.4 100 43.3 UNIX 7.8 Total 100 100 Healthcare Linux 9.9 100 2.3 Other 0.4 74.1 UNIX 13.2 0.0 4.2 0.6 0.0 2.006 2.008 10.7 6.1 9.3 Windows 85.8 100 12.3 100 100 .6 0.2 Total 100 Transportation Linux 12.3 4.4 66.2 7.5 85.6 Other 0.4 0.2 Windows 67.5 100 2.1 UNIX 10.3 Total 100 Utilities Linux 7.4 77.0 6.6 52.5 23.2 73.4 Windows 84.0 100 22.6 0.8 65.7 0.3 88.2 Total 100 Retail Trade Linux 7.2 Other 0.0 7.2 100 14.3 Other 0.6 73.1 UNIX 10.2 12.7 83.2 Other 0.5 Total 100 Wholesale trade Linux 15.0 Other 0.1 0.0 UNIX 3.1 11.4 0.2 Total 100 100 Communications Linux 21.8 0.4 79.0 12.2 0.8 0.2 0.9 0.4 Windows 77.9 78.1 100 20.5 72.2 Windows 81.2 100 36. and Construction Linux 7. Mining.2 0.3 100 32.0 5.9 9.5 Windows 86.3 83.8 84.007 Agriculture.3 25.9 100 18.Annex XX Percentage share of different OS installed across verticals 2.7 Total 100 100 Government Linux 12.3 Windows 77.0 6.5 100 12.7 0.5 100 22.3 6.007 9.0 6.0 UNIX 5.2 0.1 67.8 Total 100 Services Linux 16.3 0.0 Windows 83.0 9.5 10.2 0.

Annex XXI Percentage share of Processor sold in each vertical
2,006 2,007 Agriculture, Mining, and Construction IA64 1 0 Other 0 0 RISC 6 5 x86 93 95 Total 100 100 Communications IA64 1 2 RISC 8 7 x86 91 91 Total 100 100 Discrete Manufacturing IA64 Other RISC x86 Total Education IA64 RISC x86 Total Financial Services IA64 Other RISC x86 Total Government IA64 RISC x86 Total Healthcare IA64 RISC x86 Total 2,008 1 0 4 95 100 1 8 91 100 2,006 Process Manufacturing IA64 0 Other 0 RISC 6 x86 94 Total 100 Retail Trade IA64 0 Other 0 RISC 3 x86 96 Total 100 Services IA64 0 Other 0 RISC 5 x86 94 Total 100 Transportation IA64 0 Other 0 RISC 3 x86 97 Total 100 Utilities IA64 0 Other 0 RISC 2 x86 98 Total 100 Wholesale trade IA64 1 Other 0 RISC 3 x86 96 Total 100 2,007 1 0 5 94 100 0 0 3 97 100 1 0 5 94 100 0 0 2 98 100 0 0 1 99 100 2 0 5 93 100 2,008 1 0 3 95 100 1 0 3 96 100 2 0 4 94 100 0 0 1 98 100 0 0 2 98 100 1 0 4 95 100

0 0 4 96 100 0 5 95 100

0 0 4 96 100 0 4 96 100

1 0 4 95 100 0 1 98 100

1 0 6 93 100 1 8 91 100 2 12 87 100

1 0 6 93 100 1 5 94 100 5 9 86 100

1 0 5 94 100 1 3 97 100 5 4 91 100

101

Annex XXII
Distribution of OS installed across verticals by top three Vendors (2006-2008) Vendor Hewlett-Packard Vertical Agriculture, Mining, and Construction Communications Discrete Manufacturing Education Financial Services Government Healthcare Process Manufacturing Retail Trade Services Transportation Utilities Wholesale Trade Linux 0.3 19.8 4.6 2.5 13.5 10.8 2.5 2.7 1.6 21.2 3.3 1.9 15.3 100.0 0.1 19.8 7.5 8.3 16.1 5.0 0.0 1.9 1.6 21.7 0.1 1.4 16.5 100.0 0.0 53.8 2.2 4.2 12.2 5.1 0.3 0.7 1.1 17.7 0.4 0.1 2.3 100.0 Other 0.0 26.9 0.0 0.0 66.8 0.0 0.0 0.0 0.0 5.9 0.0 0.0 0.5 100.0 0.1 10.2 2.0 0.0 54.5 2.8 0.0 0.9 1.5 24.7 0.2 0.9 2.2 100.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 UNIX 0.1 25.4 1.1 0.9 31.1 2.2 11.0 4.2 0.7 12.5 0.7 0.4 9.6 100.0 0.8 18.4 3.9 0.3 26.1 14.6 1.7 4.6 1.6 16.9 1.4 1.3 8.4 100.0 0.5 19.2 6.7 5.7 20.7 11.2 2.1 3.9 2.9 16.1 0.4 1.7 8.9 100.0 Windows 0.6 13.9 9.2 5.0 18.4 9.2 3.0 4.8 4.9 16.3 3.6 2.7 8.4 100.0 0.5 12.8 8.0 2.8 23.6 10.4 0.5 4.3 5.2 16.6 3.3 2.6 9.2 100.0 0.0 46.2 0.0 0.0 32.9 0.0 0.0 1.4 0.0 18.6 0.0 0.0 0.8 100.0

Hewlett-Packard Total Agriculture, Mining, and IBM Construction Communications Discrete Manufacturing Education Financial Services Government Healthcare Process Manufacturing Retail Trade Services Transportation Utilities Wholesale Trade IBM Total Agriculture, Mining, and Sun Microsystems Construction Communications Discrete Manufacturing Education Financial Services Government Healthcare Process Manufacturing Retail Trade Services Transportation Utilities Wholesale Trade Sun Microsystems Total 102

Acknowledgements

The contribution of Openmainframe.org and Indicus Analytics in instigating and shaping this report is gratefully acknowledged. We thank Jeff Gould, Bibek Debroy, S.L. Rao, Mahesh Uppal, Rajiv Kumar, Laveesh Bhandari, Payal Malik, Ramrao Mundhe for their comments and advice at various stages. The contribution of Tekplus in partly handling our primary survey and Gartner for providing us with the data and resolving queries is appreciated. We also thank all the organizations who never flinched at our demand for data and interviews despite other exigencies. The report would not have achieved its current format without the able and expert assistance of Anil Gupta. We are deeply indebted to him.

103

104 .