SAP With Oracle Real Application Clusters 11g Release 2 (RAC) on Microsoft Windows 2008 Release 2
Oracle RAC on SAP Environment
Transcript of Oracle RAC on SAP Environment
Oracle9i RACfor SAP Customers
Agenda
Driving Forces Oracle9i RAC Architecture Oracle9i RAC Scalability Oracle9i RAC High Availability Oracle9i RAC and SAP’s MCOD Oracle9i RAC & Low Cost Technologies Oracle9i RAC Release Strategy
Driving ForcesDriving Forces
Driving Forces
Increased competition– Customers do not tolerate downtime– Management needs real-time information
System evolution– Growing amount of data– Changing workloads– Increasing complexity of IT landscapes– Increasing interaction between different systems
Driving Forces
Economic slowdown– Decreasing earnings– Need to improve margins– Less money available for investments
The Problem
More IT investments Less IT investments
IncreasedCompetition
SystemEvolution
EconomicSlowdown
The Solution
Get most out of existing resources– Opimize resource usage– Implement high availability solutions– Minimize administration costs
Protect existing, minimize new investments– Look for modularity and scalability of system
components– Look for low cost technologies
Simplify IT landscape (“consolidation”)
The Solution
For several years, consolidation was the only strategy to reduce costs.
Modularity, scalability and low cost technologies require distribution.
This means, that we need a new concept of systems design: Consolidated and integrated systems should be distributed to cheap and standardized components.
Oracle9i RACArchitectureOracle9i RACArchitecture
Standard Oracle Architecture
Instance
Database
Shared Nothing Architecture
DatabaseInstance
1
DatabaseInstance
2
DatabaseInstance
3
Table A
Table B
Table C
Shared Disk Architecture
DatabaseInstance
1
DatabaseInstance
2
DatabaseInstance
3
Table A
Table B
Table C
RAC Architecure
DatabaseInstance
1
DatabaseInstance
2
DatabaseInstance
3
Table A
Table B
Table C
Mirrored DiskSubsystem
Hig
h S
peed
In
terc
on
nect
CacheFusion
Oracle9i RACScalability
Oracle9i RACScalability
RAC Scalability
In the past, clustered databases (OPS)scaled well for specific types of applications:– Data Warehouse– Parallel-enabled OLTP
RAC with Cache Fusion delivers transparent scalability to all types of applications(including SAP applications)
RAC Scalability and SAP
In the past, the only way to scale the database server was to replace a small system by a larger system (“scale up”)
Oracle9i RAC provides an other option:add more small systems (“scale out”)
Benefits:– Protection of existing investments– Less new investments
mySAP.com Scalability
Presentation Application Database
mySAP.com Scalability
Presentation Application Database
mySAP.com Scalability
Presentation Application Database
mySAP.com/RAC Scalability
Presentation Application Database
mySAP.com/RAC Scalability
Presentation Application Database
mySAP.com/RAC Scalability
Presentation Application Database
Parallel SD Benchmark
Oracle9i RAC running on HP (Compaq) Tru64 3-tier system Finished: December 2001 Certified: June 2002
(2002029, 2002030, 2002031) Goal: Prove scalability with max. CPU
utilization
Parallel SD Benchmark
1 node 2 nodes 4 nodes0
2.000
4.000
6.000
8.000
10.000
12.000
Scalability: 1.8 Scalability: 1.8
Oracle9i RACHigh
Availability
Oracle9i RACHigh
Availability
Parallel Workload Study
Oracle9i RAC running on Windows 2000 2-tier systems Finished April 2002 Not intended for certification Goal: Prove scalability under conditions as
close as possible to real world environments(CPU util. between 33% and 70%)
RAC Scalability
100 SAP SD Users+ Oracle Instance 1
100 SAP SD Users+ Oracle Instance 3
100 SAP SD Users+ Oracle Instance 4
100 SAP SD Users+ Oracle Instance 2
RAC Scalability + High Availability
133 SAP SD Users+ Oracle Instance 1
134 SAP SD Users+ Oracle Instance 3
100 SAP SD Users+ Oracle Instance 4
133 SAP SD Users+ Oracle Instance 2
Parallel Workload Study
UsersUsers Throughput(Dsteps/Hour)Throughput
(Dsteps/Hour) ScalingScaling
100 SD100 SD 35,36435,364
ConfigurationConfiguration
1 node1 node
200 SD200 SD 70,32070,320 1.991.992 nodes2 nodes
300 SD300 SD 103,482103,482 2.932.933 nodes3 nodes
400 SD400 SD 133,840133,840 3.783.784 nodes4 nodes
Phase 1: Scalability
400 SD400 SD 124,812124,8123 nodes3 nodes
Phase 2: High Availability
Parallel Workload Study
0%
50%
100%
150%
200%
250%
300%
350%
400%
100Users
1 Node
200Users
2 Nodes
300Users
3 Nodes
400Users
4 Nodes
400Users
3 Nodes
Total Throughput
Throughput/Node
Oracle9i RAC& SAP’s MCODOracle9i RAC
& SAP’s MCOD
MCOD
SAP requires more and more databases for their different modules.
The SAP modules in a mySAP.com landscape are not independent, e.g. SAP SD, SAP CRM and BW interact and share data.
To guarantee the required consistency within all these databases, SAP has developed MCOD (“Multiple Components in One Database”)
Typical mySAP.com Landscape
OracleInstance
1
OracleInstance
2
OracleInstance
3
OracleInstance
4
SAP R/3Instance
1
SAP R/3Instance
2
SAP CRMInstance
1
SAP BWInstance
1
mySAP.com and MCOD
OracleInstance
1
SAP R/3Instance
1
SAP R/3Instance
2
SAP CRMInstance
3
OracleInstance
2
SAP BWInstance
4
mySAP.com, MCOD & RAC
OracleInstance
1
OracleInstance
2
OracleInstance
3
OracleInstance
4
SAP R/3Instance
1
SAP R/3Instance
2
SAP CRMInstance
1
SAP BWInstance
1
mySAP.com, MCOD & RAC
With Oracle 9i RAC, nodes can be optimally customized for dedicated workloads(e.g. CRM, HR, SD, Retail, etc.).
With Oracle 9i RAC, OLTP, BW andbatch-centric modules can be fine tuned without affecting each other.
Only running all related SAP modules inone database guarantees the consistency, especially for backups.
Oracle9i RAC& Low-Cost
Technologies
Oracle9i RAC& Low-Cost
Technologies
Blades
Up to 600 CPUsper 19“ rack
Decreased spacerequirements: up to 7 times lessspace compared to classic servers
Decreased power requirements:up to 5 times less power consumption
Decreased cooling requirements: up to 5 times less cooling
Decreased price per CPU: up to 20 times less $/CPU compared to a classic 32-way server
Blades: Expected Savings
Company 1Company 1$500,000 to $1,000,000 per rack$500,000 to $1,000,000 per rack
Company 2Company 260% per server,90% of data center space (15m2 vs. 156m2)60% per server,90% of data center space (15m2 vs. 156m2)
Company 3Company 360% of data center space (19,000m2 vs. 48,000m2)= € 7,000,00060% of data center space (19,000m2 vs. 48,000m2)= € 7,000,000
Company 4Company 4reduction from $54 to $31 per userreduction from $54 to $31 per user
Blades & Oracle9i RAC
Blades have a high potential to cut cost Oracle runs on blades 800 SD users with Oracle on a 2-way,
Intel PIII, 800MHz, blade With Oracle9i Real Application Clusters
(RAC), blades can be clustered as instances of one database because of Oracle‘s shared disk architecture
Only Oracle 9i RAC can run on more than one blade with SAP
Grid Computing
Without Grid, systems have dedicated tasks. Each system has to be sized for worst case
peak load of his task. Under normal conditions, systems run over
hours with low load. Potential CPU power is wasted, because
unused.
Grid Computing
Example 1: Retail Within the normal business
hours the system is under low load.
After business hours POS upload starts, new batch calculation starts, data collection and transfer to the BW systems starts. The system is now, but only a view hours, under the load it had to been sized for.
Retail
idle
Retail
idle
~ 50% idle
Grid Computing
Example 2: CRM Within the normal business
hours, the system is under load it was sized for.
After business hours it runs with low load till the next business day.
CRM
CRM
idle
idle
~ 50% idle
Grid Computing
Grid to minimize unused resources. Because different systems have different
resources requirements at different points in time, free resources can be shared.
With Oracle Real Application Clusters (RAC),additional database nodes can be added up on demand.Example: Remove a database node from the CRM system and assign it to the Retail system for POS uploads.
Grid Computing
~ 50% idle ~ 25% idle
Retail
idle
Retail
idle
CRM
idle
CRM
idle
idle
Retail
CRM
idle
CRM
Retail
Centralized Storage
Storage Area Network (SAN) or Network-attached Storage (NAS)
Separates storage from the traditional server and puts it on special appliances
Provides a common storage pool that is highly scalable and flexible
Centralized Storage
Centralized storage matches the capabilities offered by blades and RAC:– Small computing units without local disk– Shared disk storage (RAC)
No Oracle Cluster File System required, if NAS is used
Linux
What is it?– Open-source operating system.– Supported by many hardware vendors.– Supported by many software vendors.– Increasing market share as server
operating system.
Linux
Why can it save money?– Low-cost operating system.– Low-cost third-party software.– Homogeneous SW environment for
heterogeneous HW (e.g. Linux on Blades or nearly any Intel based HW).
– Low-cost training.
Source: IDC White Paper “Maximizing the Business Value of EnterpriseDatabase Applications on a Linux Platform” (2002)
Linux Today Linux scales very well in SMP systems up
to 4 CPU’s. Linux on 8 CPU’s is still competitive, but between
4way and 8way systems the price per CPU increases significantly.
For SMP systems with more than 8 CPU’s, classic Unix systems are the best choice.
With Oracle Real Application Clusters (RAC),small 4 or 8way systems can be clustered to cross the today’s Linux limitations.
Commodity, inexpensive 4way Intel boxes, clustered with Oracle 9i RAC, help to reduce TCO.
Linux
Oracle has a strong commitment to support Linux.
Oracle 9i RAC for SAP certification on Linux has already been started by HP, Dell, Fujitsu-Siemens.
Oracle 9i RAC for SAP will be certified for RedHat and Suse.
Release StrategyRelease Strategy
Release Strategy
SAP has– agreed to adopt, distribute and support Oracle9i
RAC (May 2002)– certified a series of SD Parallel benchmarks
using Oracle9i RAC on HP (Compaq) Tru64 Unix (May 2002).
RAC will be generally available for customers on HP (Compaq) Tru64 Unix within 6 months.
Release Strategy
Benchmark Pilot Customer(s) AvailabilityTrue64
Linux Benchmark Pilot Customer(s) Availability
Benchmark Pilot Customer(s) AvailabilityAIX
HP-UX
Solaris
Benchmark Pilot Customer(s) Avail
Benchmark Pilot Customer(s) Avail
Benchmark Pilot Customer(s) AvailWindows