You are on page 1of 4

dell enterprise technology center series

Implementation Study:
Dell IT Scales Supply
Chain Management
with Oracle RAC 10g
By Dave Jaffe, Ph.D.
When the expensive proprietary servers running the Dell supply
Todd Muirhead

Tiong Tey
chain management systems had reached their limits, the Dell
Raveendra Avutu IT group migrated to cost-effective, standards-based Dell ™
PowerEdge™ servers running Oracle® Real Application Clusters 10g.
This architecture helped enhance database performance while
providing scalability for future growth.

T
he Dell supply chain management (SCM) database implemented this type of system for Dell operations at mul-
systems handle key business functions that support tiple locations around the world. Using the same processes
worldwide manufacturing operations, including the for disaster recovery, backup, and monitoring across all
efficient Dell inventory management model and fast, direct Dell operations enables Dell IT to take advantage of a cost-
delivery of computers, accessories, parts, and supplies. These effective and readily supported deployment model. This
systems must be designed for reliability and cost-effective scal- article describes how Dell IT migrated its critical SCM appli-
ability: a failure can cost thousands of dollars per minute in cations from proprietary UNIX-based servers to industry-
Related Categories:
factory downtime, and the SCM systems must be able to standard Dell hardware.
Database handle increasing workloads as the company grows.
Dell Enterprise When Dell was a smaller company than it is now, before Dell supply chain management
Technology Center the development of powerful, industry-standard servers of The Dell SCM system must handle an enormous number of
Implementation study the type that Dell manufactures, the Dell IT group ran its SCM transactions and pieces of information, and includes multi-
database applications on large, expensive, proprietary serv- ple core components necessary to keep operations running
Oracle
ers based on the UNIX® OS. However, as the company grew, smoothly:
Scalable enterprise
servers lacking the necessary capacity had to be replaced
Visit www.dell.com/powersolutions
with even larger, more powerful servers. And because the • Configuration management: The configuration manage-
for the complete category index.
servers were not redundant, updating a single server often ment component manages over 1 million Dell part num-
required shutting down entire systems. bers per year across approximately 200 product families,
The increased performance of industry-standard Dell and over 2 million bills of materials (BOMs) per year.
PowerEdge servers, however, has enabled Dell IT to create BOMs listing component part numbers are created for
cost-effective, highly scalable systems using Oracle Real manufacturing facilities to build assemblies and sub­
Application Clusters (RAC) 10g. By sharing a large database assemblies for Dell products.
across multiple PowerEdge servers, Dell IT can easily deploy • Procurement: The procurement component manages nearly
additional low-cost servers when necessary to handle 1.8 million purchase order lines per year from more than
increased workloads rather than buying additional large, 5,000 suppliers worldwide. To streamline the procurement
expensive, proprietary UNIX-based servers. Dell IT has process, Dell uses an automated application that includes

DELL POWER SOLUTIONS | May 2007 Reprinted from Dell Power Solutions, May 2007. Copyright © 2007 Dell Inc. All rights reserved.
50
managed the production database on Oracle
Database 8.0.6. The disaster recovery server
Production data center had 12 processors at 336 MHz and 6 GB of
memory. Arrays running the EMC® Symmetrix®
Sun Cluster platform provided disk storage.
To make the system highly available, Dell IT
Domain Name System (DNS) alias for failover

used Sun Cluster to cluster the primary and sec-


ondary servers in an active/passive configura-
tion connected to shared disk storage, enabling
Primary Secondary the database to fail over to the passive node if
Sun E���� Sun E���� EMC Symmetrix
the active one failed. The failover was initiated

EMC SRDF
when the passive node no longer detected a
heartbeat signal from the active node. The
failover disconnected all users, who would then
Disaster recovery data center
have to reconnect to the newly active server.

Current system: Oracle RAC


on Dell PowerEdge servers
By 2005, the performance of the proprietary
Disaster recovery
system was becoming a concern, with many of
Sun E���� EMC Symmetrix
the batch processes taking a long time to
complete—the end-of-quarter batch process
job, for example, could take up to 31 hours. In
addition, the Oracle Database version was out-
Figure 1. Previous Dell supply chain management system based on proprietary UNIX-based servers dated and unsupported, and because it did not
support Oracle RAC, Dell IT could not add capac-
workflow approvals and vendor communica- In addition to these order-related transac- ity by horizontal scaling. To continue using this
tion and enables services such as defective tions, the SCM system also runs several other system would have required a large investment
part replacement. batch process jobs to aggregate data each week, in upgrading these large, expensive servers.
• Cost: The cost component runs mostly in month, or quarter. Instead, Dell decided to migrate to a cost-
batch mode to calculate the costs to Dell for In the North America region, the Oracle effective, industry standards–based platform,
all BOMs. These batch jobs run weekly, Database application for SCM consists of
monthly, and quarterly, with each job aggre- approximately 3,000 database objects (func-
gating total material costs. tions, packages, procedures, triggers, tables,
• Inventory: The inventory component man-
ages more than 3 million inventory move-
and views). This SCM system is supported by
six PowerEdge 2650 application servers, five
TALK BACK
ments daily from stock rooms to factory internally developed Web-based applications,
floors across all Dell sites, along with the more than 50 system-to-system integrations,
corresponding 3 million messages transmit- approximately 125 batch jobs, and approxi-
Tell us how the Dell Enterprise Tech-
ted to different systems for reporting, analy- mately 500 user interfaces.
sis, and factory scheduling. nology Center can help your organiza-
• Accounts payable: The accounts payable Previous system: Proprietary tion better simplify, utilize, and scale
component handles approximately 15,000 UNIX-based servers
enterprise solutions and platforms.
items per day, including payments to Dell The previous Dell SCM system used Sun E6000-
suppliers, invoices, and receipts. Vendor class UNIX-based servers running the Sun Send your feedback and ideas to
information includes vendor ID number, Solaris 8 OS and Sun Cluster 2.2 (see Figure 1). enterprise_techcenter@dell.com.
location, negotiated terms, and contact The primary and secondary servers, each with
information. 16 processors at 336 MHz and 11 GB of memory,

Reprinted from Dell Power Solutions, May 2007. Copyright © 2007 Dell Inc. All rights reserved. www.dell.com/powersolutions 51
dell enterprise technology center series

apply them, which allows the database to stay


in sync and helps prevent physical corruptions
from being copied to the disaster recovery data-
Production data center
base. The second cluster uses SRDF to ship
Primary Oracle RAC 10g cluster changed blocks at the storage level in real time
from the primary site to the disaster recovery
site. This method is fast but does not guard
EMC Symmetrix against data corruption. The combination of
Data Guard and SRDF provides both fast and

EMC SRDF
Oracle Data Guard redo log shipping

secure data replication.


Dell IT also uses several other programs to
Disaster recovery data center provide key backup, load balancing, and admin-
istration capabilities:
EMC SRDF disaster recovery Oracle RAC 10g cluster

• Oracle Recovery Manager: Dell IT uses this


program to back up the primary database
EMC Symmetrix and archive logs. Two full (hot) backups are
made to tape weekly through the EMC
NetWorker™ interface; archive logs are
Oracle Data Guard disaster recovery Oracle RAC 10g cluster backed up hourly to enable Dell IT to recover
the database to a specific point in time if
necessary. Dell IT can also back up the data-
base from the disaster recovery site rather
EMC Symmetrix
than taxing the primary nodes.
• Oracle Database 10g Services: This software
provides load balancing for each cluster—for
example, Dell IT can have three nodes

Figure 2. Current Dell supply chain management system based on Oracle RAC 10g and Dell handle online users and three nodes handle
PowerEdge servers batch processing.
• Oracle Grid Control: This software provides
replacing the Sun servers with Oracle RAC 10g is seamless and transparent to users, users do monitoring and management capabilities
clusters of Dell PowerEdge 6650 servers, each not have to reconnect, as they did in the previ- and helps simplify many day-to-day data-
with four Intel® Xeon® processors (see Figure 2). ous system. base administration tasks; it also allows
The environment includes a primary cluster and One of the disaster recovery clusters uses administrators to set thresholds for different
two disaster recovery clusters. While the previ- Oracle Data Guard software, and the other uses events to create trouble tickets through the
ous system could not scale beyond one physical EMC Symmetrix Remote Data Facility (SRDF®) problem management system.
server, the current system allows Dell IT to easily software. The first cluster uses the Data Guard
add servers to the clusters to handle increased Broker component to ship logs in real time from Dell IT has implemented these clusters at mul-
workloads. The multi-node Oracle RAC architec- the primary site to the disaster recovery site and tiple locations around the world (see Figure 3).
ture can also provide higher I/O throughput
than the previous system because of the

“Standardizing on cost-effective Oracle RAC 10g


increased number of available interfaces.
The Oracle RAC 10g architecture includes
built-in high-availability features. All Oracle RAC
nodes and instances share the same physical
clusters of PowerEdge servers with Dell/EMC stor-
database; if any of these experience problems, age helped Dell IT deploy the systems globally in
the users and connections can fail over to the
other nodes or instances. Because this failover just eight months.”
DELL POWER SOLUTIONS | May 2007 Reprinted from Dell Power Solutions, May 2007. Copyright © 2007 Dell Inc. All rights reserved.
52
Standardizing on cost-effective Oracle RAC 10g data extraction for all material movements SCM systems to provide real-time information
clusters of PowerEdge servers with Dell/EMC stor- transactions, which dropped from almost about key business functions. Running these sys-
age, with similar processes for deployment, disas- 5 hours to just 35 minutes for an 88 percent tems on Oracle RAC 10g clusters of industry-
ter recovery, and backup, helped Dell IT deploy improvement, while the time for the entire standard Dell PowerEdge servers enables Dell IT
the systems globally in just eight months. end-of-quarter jobs processing decreased to scale them efficiently and cost-effectively to
from 31 hours to 23 hours. handle increased workloads. By moving the sys-
Performance increases tems to Dell servers when it did, Dell IT avoided
Dell engineers measured database perfor- Cost-effective, scalable supply chain significant additional expenditures for proprietary
mance for both the previous and current SCM management UNIX-based servers, enhanced performance, and
systems. Figure 4 shows the 10 longest end- Supply chain management is essential to Dell provided a clear path for future growth.
of-month and end-of-quarter transactions. The operations around the world, with both factory
most significant time savings occurred in the operations and internal systems dependent on Dave Jaffe, Ph.D., is a senior consultant on the
Dell Enterprise Technology Center team. He has
a B.S. in Chemistry from Yale University and a
Region Primary site Disaster recovery site Ph.D. in Chemistry from the University of
North America Six-node Oracle RAC 10g cluster Two sets of six-node Oracle RAC 10g clusters California, San Diego.

China Three-node Oracle RAC 10g cluster Three-node Oracle RAC 10g cluster
Todd Muirhead is a senior engineering consultant
Brazil Two-node Oracle RAC 10g cluster Two-node Oracle RAC 10g cluster
on the Dell Enterprise Technology Center team.
Europe Two-node Oracle RAC 10g cluster Two-node Oracle RAC 10g cluster Todd has a B.A. in Computer Science from the

Malaysia Two sets of two-node Oracle RAC 10g clusters Two sets of two-node Oracle RAC 10g clusters University of North Texas and is Microsoft Certified
Systems Engineer + Internet (MCSE+I) certified.

Figure 3. Global deployment of Dell supply chain management systems


Tiong Tey is an IT strategist on the Dell World
Wide Procurement IT team. Tiong has a B.S. and
Transaction Time on previous Time on current Performance gain a master’s degree in Computer Science from
system system Wichita State University.
(hours:minutes) (hours:minutes)

Entire end-of-quarter jobs 31:00 23:00 26% Raveendra Avutu is an IT manager in the Global
processing
Data Management group supporting databases
Cost-per-order calculation 8:52 5:40 36%
for Dell IT. Raveendra has a master’s degree in
Quarterly BOM calculation 7:30 1:56 74% Computer Science from Oklahoma State University.

Entire end-of-month jobs 6:00 4:00 33%


processing

New product material costs 5:56 2:17 62%


calculation

Data summary for all levels 5:33 1:58 65%


of order details

Data extraction for all material 4:49 0:35 88%


movements transactions
QUICK LINKs
Costs roll-up (materials, 3:18 3:12 3%
royalties, and transportation)
Dell Enterprise Technology Center:
Material management transac- 3:00 0:43 76% www.dell.com/techcenter
tions creation based on inventory
transactions Dell and Oracle Database 10g:
www.dell.com/oracle10g
Financial journal entries creation 2:37 0:40 75%
Oracle and Dell:
www.dell.com/oracle
Figure 4. Performance gains for 10 longest database transactions in Dell supply chain
management systems

Reprinted from Dell Power Solutions, May 2007. Copyright © 2007 Dell Inc. All rights reserved. www.dell.com/powersolutions 53

You might also like