You are on page 1of 30

BI Archiving by usage of PBS-NLS

Marjolijn Berends (St-Addons) en Peter Tonies (NXP) 7 December 2011

Agenda
NXP System/Application Landscape Need for BI Archiving ST-Addons: Near Line Storage (NLS) Solution Proof of Concept BI Archiving

Next steps and Summary

NXP Semiconductors
NXP Semiconductors N.V. (NASDAQ: NXPI) provides High Performance Mixed Signal and Standard Product solutions that leverage its leading RF, Analog, Power Management, Interface, Security and Digital Processing expertise. Our innovations are used in a wide range of automotive, identification, wireless infrastructure, lighting, industrial, mobile, consumer and computing applications. A global semiconductor company with operations in more than 25 countries, NXP posted revenue of $4.4 billion in 2010.

Key applications (1/2)


Wireless infra
Lighting Industrial Mobile

Wireless base stations Point-to-point

Lighting drivers (CFL, LED) Lighting networks Backlighting

Smart grid White goods Home / building automation Power supplies

Mobile devices Portable power supplies

CATV infrastructure
Broadcasting

Personal health
Chargers

Key applications (2/2)


Automotive Identification Consumer Computing

In-vehicle networking Car access & immobilizers Car entertainment Solid State Lighting Telematics Lighting

Secure identity Secure transactions

TV Satellite, Cable, Terrestrial and IP set-top boxes Satellite outdoor units

Tablet PCs Note- /Netbooks Desktops Power supplies. Monitors and peripherals

Tagging & authentication

Agenda
NXP System/Application Landscape Need for BI Archiving ST-Addons: Near Line Storage (NLS) Solution Proof of Concept BI Archiving

Next steps and Summary

BI application landscape (green=used)


SAP BW Portal

Bex analyzer

CHAMP (Netweaver)

Some SAP BW characteristics

Number per June 2011 (+/-) Active users 1200

Active queries used


Infoproviders (cubes/DSO/info objects) Database size Nr. Of records loaded per week

2500
> 1000 2.2 Terabyte 10 billion

Agenda
NXP System/Application Landscape Need for BI Archiving ST-Addons: Near Line Storage (NLS) Solution Proof of Concept BI Archiving

Next steps and Summary

Why would we need to archive in SAP BW?


1.

BW Accelerator requires that data growth is controlled -> this can be achieved by regularly archiving data

2.

Reduce size of online database, which leads to:


Improved reporting response times (reading from archive is faster then from online) Improved data loading times -> more predictable times for data availability Reduced costs in case we can store the archived data on a lower cost storage medium OR in case the archived data uses less space due to compression

What if not?
Strong growth of BW database leads to operational issues and complexity (adding storage) Performance decreasing over time Inefficient usage of storage

History regarding BI archiving at NXP


Archiving policy defined Until December 2010: used standard SAP archiving functionality Two areas piloted and archived:
Inventory management Profit Center Accounting

Main disadvantages of standard SAP archiving functionality:


To read archived data it must be first reloaded into BW (Complex, timeconsuming) This implies that we can only archive data once we hardly ever need to access it again No compression so hardly any reduction of storage costs

Other possibilities for Archiving -> Near Line Storage solutions from Development Partners

Agenda
NXP System/Application Landscape Need for BI Archiving ST-Addons: Near Line Storage (NLS) Solution Proof of Concept BI Archiving

Next steps and Summary

13

NLS Solutions for SAP BW:


Data Archiving Process DAP

Column-based Architecture
Relational Database
c 1 r1 r2 r3 r4 r5 c 2 c 3 c 4 c 5 c 6 c 7 c 8 c 9

Data is stored horizontally in rows Reporting is very I/O-intensive No effective query access

SYBASE IQ
c 1 r1 r2 r3 r4 r5 c 2 c 3 c 4 c 5 c 6 c 7 c 8 c 9

Data is stored vertically in columns

Each column acts as index for its data


Reporting: significantly reduced I/O rate Effective query access options
Source: Sybase Inc.

NLS Solutions for SAP BW:


PBS MultiProvider Integration for BW 7.0

Agenda
NXP System/Application Landscape Need for BI Archiving ST-Addons: Near Line Storage (NLS) Solution Proof of Concept (PoC) BI Archiving
Why PoC? Scope PoC Expectations PoC Results PoC Business Case

Next steps and Summary

17

Reasons for executing PoC archiving with PBS


Need solution that can provide direct access to archived data via the query, meaning that we can shorten time to archive -> PBS software
Main expected benefits:
Complete integration between online SAP data and archived SAP data Guaranteed flexible access to our SAP data in the future Impact on SAP users is zero by using the SAP standard reports/transactions Discussion what do I keep online moves from retention period to is this document closed? Archive it a.s.a.p. (Information Lifecycle Mngt) Reduces the load on SAP BW system and reduce TCO

Disadvantage:
Investment required (both license and hardware)

Proof Of Concept costs: appr. 35 kUSD

Proof of Concept (PoC) BI Archiving -> Why PoC?

Scope of PoC Archiving with PBS


Datamodel PCA (which is part of POC PBS NLS)

NLS IQ

SFF_C63: PCA High Level

SFF_C64: PCA Details

SFF_O66: PCA Container Line Items

SFF_O60

SFF_O63

SFF_O64: PCA Actuals Line Items CLASS

SFF_O61: PCA Act Items SPEED

Proof of Concept (PoC) BI Archiving -> Scope PoC

19

Acceptance criteria PoC Archiving with PBS


Criteria Expected result

Data Quality / Data content consistency


Possibility to archive DSO data Possibility to archive cube data Possibility to report directly on archived data Response time when reading from archived data at least 2 times faster then from online db Database size of archived data is 10% or less from original db size Data reload time

Yes
Yes Yes Yes In most cases Yes 1 day

Proof of Concept (PoC) BI Archiving -> Expectations

20

Acceptance criteria PoC Archiving with PBS


Criteria Expected result Actual result

Data Quality / Data content consistency


Possibility to archive DSO data Possibility to archive cube data Possibility to report directly on archived data Response time when reading from archived data at least 2 times faster then from online db Database size of archived data is 10% or less from original db size Data reload time

Yes
Yes Yes Yes In most cases Yes 1 day

Yes
Yes Yes Yes Yes No 4 a 5 hours

Proof of Concept (PoC) BI Archiving -> Results I

21

Query response times NLS vs Online database (lower value means faster response time)
Response time in Seconds
250 218 200 155 150 96 65 50 37 9 0 Query 1, NLS data (3.016.771) Query 1, Online data (2.708.379) Query 2, NLS data Query 2, Online Query 3, NLS data (1.962.052) data (8.698.215) (6.980.163) Tested query ( avg. selected records ) Query 3, Online data (5.390.737) 32 13 6 1 before archiving 68 After archiving

100

92

Query 1 & 2 are real-world PCA querys Query 3 is a test query aggregating 1 keyfigure of a large amount PCA data on year level. Query 3 online data performance increase also due to other factors (upgrade Oracle 11) Correctness of output data all querys verified

Proof of Concept (PoC) BI Archiving -> Results II

22

Archiving & reloading times


Monthly archiving of total PCA model will take approx. 4.5 hours Monthly archiving run (6.5mln records) of the biggest data container, the PCA integration layer (total 160mln records), takes +/- 1.5 hours. Average archiving speed: 75k records / minute Reloading time nearly equals original archiving time.

Write

Store

Delete

Restore

Proof of Concept (PoC) BI Archiving -> Results III

23

Sizing
Decrease online storage (by the amount of archived data) Increase NLS storage
Compression factor: less promising results after Oracle 11.2g then initially, but measurements so far show a reduction of 50%

Proof of Concept (PoC) BI Archiving -> Results IV

24

Required costs for BW archiving with PBS


One-time investment required for BW Archving
License Server Project costs Total one-time investment 76% 3% 21% 100%

Yearly fee (recurring) Maintenance fee for PBS software (17%) Server costs Production (. eur/mnth) Optional: Server costs Quality assurance/test ( eur/mnth) TAM costs Total yearly fee (recurring)

2011 13% 12% 12% 2% 39%

2012 13% 12% 12% 2% 39%

2013 13% 12% 12% 2% 39%

2014 13% 12% 12% 2% 39%

Proof of Concept (PoC) BI Archiving -> Business Case I

25

Business case PBS archiving - 1


Based on reduction of storage costs:
Assumptions
Assumption: costs for storage (in eur/mnth/Gb (including backup)) Assumption: costs for storage for archived date (in eur/mnth/Gb ) Expected yearly database growth (%) Remaining database size after archiving (in %) 1 0.75 50 50

120000 100000 80000 60000 40000 20000 0 Cost for storage (yearly) when using archiving Cost for storage (yearly) w/o archiving Net saving

2011
-20000 -40000 -60000

2012

2013

2014

Conclusion: no business case.BUT:


Proof of Concept (PoC) BI Archiving -> Business Case II
26

Business case PBS archiving - 2


The BW Accelerator is growing fast in size and is currently almost full - > archiving means that the size of the BWA can be controlled and therefore avoids an additional investment in the BWA ! Assumptions:
Yearly BWA growth (in %) Costs for additional storage BWA (in eur/Gb, one-time investment) 60 3400

900000 800000 700000 600000 500000 400000 300000 200000 100000 Cumulative net saving including avoidance of additional BWA costs 2011 -100000 2012 2013 2014 Net saving including avoidance of additional BWA costs per year Net saving

Conclusion: savings as of 2012 (earn-back time appr. 1 year) PLUS


Proof of Concept (PoC) BI Archiving -> Business Case III
27

Agenda
NXP System/Application Landscape Need for BI Archiving ST-Addons: Near Line Storage (NLS) Solution Proof of Concept BI Archiving

Next steps and Summary

28

Next steps
Decide on initial investment Arrange license for PBS together with purchasing Make PBS archiving productive
Install production server and productive license Determine order of infocubes/DSOs to archive (based on combination of size and availability of a partitioning entity) Embed in structural process

29

Summary
Archiving in BW is a must. There are two options:
Archiving solution
Standard SAP (ADK)

Pro
Cheap

Con
Investment in BWA expansion required Time-consuming and cumbersome process to read from archive plus issue when data structure changes In case no investment in BWA expansion is/can be done: certain functional areas will have to be removed from the BWA -> performance decrease for end users No compression so hardly any saving on storage costs

Remarks

PBS archiving solution

Reporting performance improvement, read directly from archive Data loading performance improvement Enables control over growth of BWA, no investment in BWA required

PBS archiving solution

Strong business case in case majority of data in BWA is suitable for archiving

30

You might also like