White PapersMany PeopleSoft Enterprise customers have the require-ment to support different environments. The ability to consolidate into a virtualized environment helps IT respond more effectively to business goals. One way to successfully accomplish this is by using logical partitions to divide the server and accommodate large workloads on one server without affecting other applications. IBM has taken another step forward by adding additional functionality to its “virtual-ized” suite of products.
IBM’s POWER5™ technology feature, Advanced POWER™ Virtualization (APV), was tested on an IBM System p5™ p570 16-way server using PeopleSoft Enterprise bench-marks. APV allows the pooling of IT resources in a way that masks the physical nature and boundaries of those re-sources from users. The goal of these tests was to demon-strate the power and functionality of this feature in a customer environment using both online and batch envi-ronments on one server.
IBM Oracle International Competency Center November 2006 IBM Advanced POWER Virtualization and Oracle s PeopleSoft Enterprise Application Functionality Tests Amy I. Falos Oracle Applications Benchmark Lead IBM ISV Business Strategy and Enablement Erik Kane Technical Consultant Team Lead IBM ISV Business Strategy and Enablement Untitled Document IBM Advanced POWER Virtualization and Oracle s PeopleSoft Enterprise Version 11/06/2006 http://www.ibm.com/support/techdocs Table of contents 2 Introduction 2 Virtualization Introduction Many PeopleSoft Enterprise customers have the require-ment to support different environments. The ability to consolidate into a virtualized environment helps IT respond more effectively to business goals. One way to successfully accomplish this is by using logical partitions to divide the server and accommodate large workloads on one server without affecting other applications. IBM has taken another step forward by adding additional functionality to its virtual-ized suite of products. 3 Highlights 4 Virtualization concepts 7 Virtual CPU 10 Virtual I/O 13 Virtual Ethernet 14 Conclusions 15 Appendix A System p and Vir-tualization IBM s POWER5" technology feature, Advanced POWER" Virtualization (APV), was tested on an IBM System p5" p570 16-way server using PeopleSoft Enterprise bench-marks. APV allows the pooling of IT resources in a way that masks the physical nature and boundaries of those re-sources from users. The goal of these tests was to demon-strate the power and functionality of this feature in a customer environment using both online and batch envi-ronments on one server. Virtualization The value of IBM s virtualization: " Reduce costs by increasing asset utilization via vir-tualized CPU, network and disk adapters. " Simplify server management and operations " Redeploy talent to manage your business, not your infrastructure " Rapidly provision new servers " Drive new levels of IT productivity with intelligent resource allocation on demand Advanced POWER Virtualization includes the following components: Micro-partitioning" " Share processors across multiple partitions " Minimum partition can be allocated only 1/10th of a processor " Maximum number of partitions is 254 " AIX 5L" V5.3, Linux *, or i5/OS"** Virtual IO server " Shared Ethernet Copyright 2006, IBM Corporation Page 2 Untitled Document IBM Advanced POWER Virtualization and Oracle s PeopleSoft Enterprise Version 11/06/2006 http://www.ibm.com/support/techdocs" Shared SCSI and Fibre Channel-attached disk subsystems " Supports AIX 5L V5.3 and Linux* partitions " Managed via HMC or IVM Partition load manager " Balances processor and memory request " Both AIX 5L V5.2 and AIX 5L V5.3 supported * Supported Linux Versions: " SuSE Linux Enterprise Server (SLES) Version 9 SP1 or later " Red Hat Enterprise Linux (RHEL) Version 3 U3 or later " Gentoo Linux " SLES Version 10 was just announced with memory add ability only Note: Not all APV features may be available depending on the distribution and version. **i5/OS - Available on 1.65 GHz System p5 570, p5 590 and p5 595 models. Highlights To demonstrate the use of APV, two PeopleSoft Enterprise applications were tested. The standard benchmark kits for HR Self Service 8.9 (HR) and Global Payroll France 8.9 (GPF) from Oracle s bench-marking group were selected. HR is an online transaction application whereby 2,000 concurrent users were run via a web browser. GPF is a heavy I/O batch application utilizing 150,000 payees. This combination was chosen to show real life situations where the use of APV can enhance availability and maximize the power of the server. Virtualized CPU This paper reviews three operational scenarios to present the capabilities of APV. Each of the scenarios demonstrated improved performance of that partition and maintained the level of run time or response time on the other partitions. " Scenario one = 2 Logical Partitions (LPAR), with HR configured as high priority. Using configura-tion options with the Hardware Management Console (HMC), we set the CPU shared pool with a higher weight. This higher weight in the LPAR configuration allows the scheduler to give priority to requests for the shared CPU pool to this partition. " Scenario two = 2 LPARs, GPF high priority. This configuration is the reverse of scenario one, with GPF having the highest priority via the weight setting. " Scenario three = 3 LPARs. This test had GPF running for about 30 min before starting up the HR run. The results demonstrate minimal impact that launching of the HR application has to the shared pool resources. GPF maintained a constant level of service while allowing the HR users to run. Virtual IO server Virtual IO server is an APV feature that allows access to disk and Ethernet adapters from a central server to all configured partitions. The fiber channel or SCSI controllers in the VIO Server can be accessed by Copyright 2006, IBM Corporation Page 3 Untitled Document IBM Advanced POWER Virtualization and Oracle s PeopleSoft Enterprise Version 11/06/2006 http://www.ibm.com/support/techdocsusing virtual SCSI controllers in the client partitions. This represents a cost savings on adapter cards and centralizes system management. This feature was tested to analyze the effect of disk IO over a VIO server and the effect on response or run times of the test partition. The test results proved that by implementing the VIO server, response and run times for both the batch and online applications can be maintained with fewer components and more manageability. Virtualization concepts Virtual consoles The POWER Hypervisor" Virtual Terminal support capability provides console access to every logical partition without a dedicated physical device . The console emulates a vt320 terminal that can be used to access the partition system using the Hardware Management Console (HMC). A partition s device tree, containing one or more nodes, can be assigned to one or more virtual terminal (vterm) client adapters. The unit address of the node is used by the partition to map the virtual device (or devices) to the operat-ing system s corresponding logical representations and notify the partition that the virtual adapter is a vterm client adapter. The node s interrupts property specifies the interrupt source number that has been assigned to the client vterm I/O adapter for receive data. Micro-partitioning Micro-partitioning enables multiple partitions to share one physical processor. The shared processor partitions are dispatched and time-sliced on the physical processors under control of the POWER Hypervisor. The physical processor is broken up into processing units, each unit represents a micro-partition , or virtual processor. Processing units The processing units are the term used in HMC to specify the processor capacity assigned to each partition. The minimum processing units is 1/10 of a processor which is specified as 0.1 processing units. For example, on a system with two physical processors, if each processing unit was one half of a physical processor (.50) then a maximum of four processing units can be assigned to a partition. In our tests we used 50% of one processor so 0.50 was our processing unit. After a partition is activated, these process-ing units become the virtual processors. See figure 1 for a pictorial view of configuring the virtual proces-sors. The pitcher represents one physical processor. It gets divided into processing units, depicted as the glasses. Each processing unit can be 1/10 up to one entire processor. The processing units then become the virtual processors when running in shared partition mode. Figure 2 shows a screen print of the configuration of a partition to show an example of one of our configu-rations. This shows how to configure the processing units and the number of virtual processors. In the shared partition mode, the number of virtual processors is used upon startup. In dedicated partition mode, the number of processing units is the number of physical processors to be activated upon partition startup. Dedicated processors are whole processors that are assigned to a single partition. If you choose to assign dedicated processors to a logical partition, at least one entire processor must be assigned to that partition. Shared and dedicated partitions can not be mixed on the same partition. If dedicated processors were being configured, the virtual processor block would be grayed out. The description here hopefully will make this process less confusing, learning how to configure and use logical partitions can be hard to grasp initially. Copyright 2006, IBM Corporation Page 4 Untitled Document IBM Advanced POWER Virtualization and Oracle s PeopleSoft Enterprise Version 11/06/2006 http://www.ibm.com/support/techdocsThe physical processoris split into processing unitsThe process unitbecomes the virtualprocessorVirtualization Configuration ConceptsVirtualization Configuration ConceptsVirtual ProcessorsPhysical ProcessorProcessing Units Figure 1, Virtualization concepts Figure 2. LPAR processor configuration detail Copyright 2006, IBM Corporation Page 5 Untitled Document IBM Advanced POWER Virtualization and Oracle s PeopleSoft Enterprise Version 11/06/2006 http://www.ibm.com/support/techdocsConfigure shared pool of virtual processors Virtual processors represent the number of concurrent operations that the operating system can use. The processing power can be conceptualized as being shared equally across these virtual processors, in one large pool. The size of the pool is defined by the number of virtual processors between the desired setting and the maximum setting in HMC not specifically assigned to a partition. The minimum value is the lowest amount required for the partition to boot. The desired number is the number that will be allocated to the partition if enough resource is available. If not enough resources are available, a lower amount will be used. The maximum value is only used as an upper limit for dynamic partitioning operations. In early tests, the desired virtual processing was set equal to maximum processing units and as a result, no shared pool was created. Selecting the optimal number of virtual processors depends on the partition workload. Some partitions benefit from a greater number of virtual processors where other partitions require greater processing power by having fewer large virtual processors. By making the number of virtual processors too small, this limits the processing capacity of an uncapped partition. If a partition with .50 processing units is configured with only one virtual processor, the partition cannot exceed one processing unit, because it can only run one job at a time, which cannot exceed one processing unit. However, if the same partition with .50 processing units was assigned two virtual processors and processing resources were available, the partition could use an additional processing unit. This is a very important point for the configuration of virtual processing. The processing capacity is usually referred to as a percentage of entitled capacity shown in the lparstat utility. The lparstat utility was used to monitor the shared pool in the tests reviewed in this paper. Capped vs. Uncapped mode Capped mode The amount of processing units is restricted to the assigned processing capacity (maxi-mum processing units). Uncapped mode The processing capacity is not restricted and that LPAR can utilize additional virtual processors when the shared pool has available resources. When a partition is running in uncapped mode, the uncapped weight, or priority, of that partition must be specified. An uncapped partition is only limited in its ability to consume cycles by the lack of online virtual proces-sors and its variable capacity weight attribute. The variable capacity weight attribute is a number be-tween 0 255, which represents the relative share of available capacity that the partition is eligible to receive. This parameter applies only to uncapped partitions. A partition s share is computed by dividing its variable capacity weight by the sum of the variable capacity weights for all uncapped partitions, resulting in the percentage of available capacity that partition is eligible to receive. Therefore, a value of 0 may be used to prevent a partition from receiving extra capacity. This is sometimes referred to as a soft cap . If multiple uncapped logical partitions require idle processing units, the managed system distributes idle processing units to the logical partitions in proportion to each logical partition s uncapped weight. The higher the uncapped weight, the more processing units the logical partition gets. All the tests detailed in this paper were run using uncapped mode. Copyright 2006, IBM Corporation Page 6 Untitled Document IBM Advanced POWER Virtualization and Oracle s PeopleSoft Enterprise Version 11/06/2006 http://www.ibm.com/support/techdocsVirtual CPU For some applications, multiple small logical partitions (LPAR)s with partial processor units can increase performance substantially over a single LPAR with whole or multiple processor resource. Tests reviewed here using one half processor did not show any gain in performance over those using one entire proces-sor for these particular applications. Customers should perform thorough testing to determine the processor units and LPAR size that is required for their particular applications best performance. See figure 3 for the testing configuration utilized in these tests. VViirrttuuaall CCPPUU TTeesstt SSeerrvveerr CCoonnffiigguurraattiioonn POWER HypervisorHR Self ServiceAIX 5L V5.3 CPU=5 entitled Memory=64GB Global PayrollAIX 5L V5.3 CPU=5 entitledMemory 64GB 128GB memory 16 1.9GHz CPU p570 VIO Server on 1LPAR GPF on 1 LPAR HR on 1 LPAR Disks attached to both LPARS IBM Systems Storage DS4800 IBM eServer System Figure 3. Virtual CPU test configuration Scenario One HR is allocated higher priority to show typical daytime activity where a payroll batch needs to be run but should not disrupt the HR online users. HR gets more CPU (4 of the shared available 6) due to higher weight in the configuration. Processing Units = .50 processors Virtual processors Weight Min:1, desired:5, max:16 (HR) Min:1, desired:16, max:32 Uncapped weight 155 Min:1, desired:5, max:16 (GPF) Min:1, desired:16, max:32 Uncapped weight 128 VIO server was not used in this configuration Note: 64GB memory on each partition Table 1. Scenario one HMC LPAR configuration Figure 4 below shows the available pool for Scenario One. HR is running the five dedicated processors in purple, while GPF is running on the five dedicated processors in green. There are six processors available in the shared pool above. During the Scenario One test, whereby HR was allocated the higher priority and allocated four more processors, as illustrated by the color coding in the shared pool of figure Copyright 2006, IBM Corporation Page 7 Untitled Document IBM Advanced POWER Virtualization and Oracle s PeopleSoft Enterprise Version 11/06/2006 http://www.ibm.com/support/techdocs3. GPF used two more processors in the shared pool which allowed for it to maintain steady perform-ance. IIBBMM eeSSeerrvveerr SSyysstteemm pp Shared Pool Scenario 1: Peoplesoft HR Self Service is given higher priority during this run HR Self ServiceGlobal Payroll FranceFigure 4. Scenario one processor shared pool configuration Scenario Two GPF batch is allocated higher priority to show batch running at night when HR Online users are still working but willing to share resource. Table 2 shows the shared pool while both applications are running. GPF is allocated more CPU (8.5 vs 7.5) due to it s higher weight in the LPAR configuration. Note that because the processing unit is set a .50 one of the shared pool processors is split between the two applications. Table 2 below shows the LPAR configuration parameters set up for these tests. This was done using the HMC graphical interface for all partitions on the same system. Processing Units = .50 processors (half of one physical processor) Virtual processors Weight Min:1, desired:5, max:16 (HR) Min:1, desired:16, max:32 Uncapped weight 128 Min:1, desired:5, max:16 (GPF) Min:1, desired:16, max:32 Uncapped weight 155 VIO server was not used in this configuration Table 2. Scenario two HMC LPAR Copyright 2006, IBM Corporation Page 8 Untitled Document IBM Advanced POWER Virtualization and Oracle s PeopleSoft Enterprise Version 11/06/2006 http://www.ibm.com/support/techdocsFigure 5 below shows the available pool for Scenario Two. HR is running the five dedicated processors in purple, while GPF is running on the five dedicated processors in green. There are six processors available in the shared pool above. During the Scenario Two test, whereby GPF was allocated the higher priority and used the shared pool to processes it s requests with three and a half more processors, as illustrated by the color coding in the shared pool of Figure 4. HR used two and a half processors in the shared pool which allowed the online users to continue without interruption. The reason we see half processors is that our processing units were configured as .50, which is one half of a processor. See table 1 for the LPAR configuration details. IIBBMM eeSSeerrvveerr SSyysstteemm pp Shared Pool Scenario 2: PeopleSoft Global Payroll France is given higher priority during this run HR Self ServiceGlobal Payroll France Figure 5. Scenario two processor shared pool configuration Scenario Three GPF batch is allocated higher priority in this scenario as well, but this server was partitioned differently. The HR database was on one LPAR and the web/application servers were on another LPAR. The GPF batch application was on the third LPAR. This shows a different load variance and since the database was isolated, the application server was highlighted. Starting the HR online user 30 minutes after the GPF batch was running shows the HR application gaining the required processors without disrupting the GPF processing. See figure 6 below. Table 3 below shows the LPAR configuration parameters set up for these tests. Note there are three partitions instead of two for this test. This was done using the HMC graphical interface for all partitions on the same system. Copyright 2006, IBM Corporation Page 9 Untitled Document IBM Advanced POWER Virtualization and Oracle s PeopleSoft Enterprise Version 11/06/2006 http://www.ibm.com/support/techdocs Processing Units = .5 processors Virtual processors Weight Min:1, desired:5, max:16 (HR app LPAR) Min:1, des:16, max:32 Uncapped weight 128 Min:1, desired:2, max:4 (HR db LPAR) Min:1, desired:2, max:4 Effectively capped due to settings but still configured as uncapped Min:1, des:5, max:16 (GPF LPAR) Min:1, des:16, max:32 Uncapped weight 155 VIO server was not used in this configuration Table 3. Scenario three HMC LPAR configuration Global Payroll France higher priority 3LPARGlobal Payroll France CPU Utilization0204060801001201 20 39 58 77 96115134153172191210229%Sys CPU%User CPUHR CPU utilization0204060801001201 20 39 58 77 96115134153172191210229%Sys CPU%User CPUGPF Virtual CPU usage0501001502002503001 16 31 46 61 76 91106121136151166181196211226% Entitlement (6)Physical CPUPhysical CPUs05101520GPF Virtual CPU usage0501001502002503001 16 31 46 61 76 91106121136151166181196211226% Entitlement (6)Physical CPUPhysical CPUs05101520HR Virtual CPU usage 0501001502002501 22 43 64 85106127148169190211232% of Entitlement (3)Physical CPUNumber of physical CP02461 1835526986103120137154171188205222239HR Virtual CPU usage 0501001502002501 22 43 64 85106127148169190211232% of Entitlement (3)Physical CPUNumber of physical CP02461 1835526986103120137154171188205222239HR DB server virtual usage0204060801001201401601 12 23 34 45 56 67 78 89 100111122% Entitlement (2)Physical CPUHR DB Physical CPU00.511.522.531 12 23 34 45 56 67 78 89100111122HR DB server virtual usage0204060801001201401601 12 23 34 45 56 67 78 89 100111122% Entitlement (2)Physical CPUHR DB Physical CPU00.511.522.531 12 23 34 45 56 67 78 89100111122Scenario 3: HMC configuration of highervirtual processors than entitlement creates the shared pool. Example of HR being started later in cycle. Figure 6. Scenario three processor shared pool configuration Virtual I/O Virtual IO Server Configuration The Virtual I/O Server (VIOS) allows sharing of physical resources between LPARs including virtual Small Computer System Interface (SCSI) and virtual networking. This allows more efficient utilization of physical resources through sharing. The VIOS provides virtual storage and shared Ethernet adapter capability to client LPARs on the system. VIOS partitions are not intended to run applications or for general user logins. Once a partition is identified as this type of partition, only the VIOS software boot image will start successfully when the partition is activated. This partition is a special type of partition which is identified as such on the first screen of the Create Logical Partitioning Wizard program. The Virtual I/O server is implemented as a customized AIX 5L partition, however, the interface to the system is abstracted using a secure shell-based command line interpreter (CLI). This partition should be Copyright 2006, IBM Corporation Page 10 Untitled Document IBM Advanced POWER Virtualization and Oracle s PeopleSoft Enterprise Version 11/06/2006 http://www.ibm.com/support/techdocsconfigured to contain physical adapter slots for shared SCSI and shared Ethernet devices for use by client partitions. These physical devices will be configured on the client partitions by using virtual adapt-ers. The partition can be configured to use dedicated processors, or processing capacity from the shared processor pool. It does not require an entire processor. It does require a minimum of 16GB disk space and 512MB of memory. VViirrttuuaall II//OO TTeesstt SSeerrvveerr CCoonnffiigguurraattiioonn POWER HypervisorHR Self ServAIX 5L V5.3 CPU=5 Mem=60GB IOSV5.3 CPU=2 Mem=4GBGlobal PayrollAIX 5L V5.3 CPU=5 Mem=64GB 16 1.9GHz CPU p570 128GB memory VIO Server on 1LPAR GPF on 1 LPAR HR on 1 LPAR Disks were attached to the VIO server IBM Systems Storage DS4800 IBM eServer System Figure 7. Virtual I/O infrastructure VIO Server Tests using Global Payroll France Figure 8 shows the server configuration used in these tests. The VIO Server tests consisted of a baseline with each application running alone on a separate partition with direct connection to the IBM System Storage" DS4800. Following that test, the DS4800 storage was attached to the VIO server and the same tests were run to measure the performance. Copyright 2006, IBM Corporation Page 11 Untitled Document Virtual IO Server Tests Virtual IO Server Tests Global Payroll FranceGlobal Payroll FranceGlobal Payroll France batch test using VIO " Achieved a 43% improvement on high volume batch with an IBM p570" VIO Disks were attached to the VIO partition and accessed using virtual IO" IO transaction throughput percentage delta is paycheck/hr62,59414443,689206Total2238Banking3832Payslip84136PayrollPaycheck/HRMinutesPaycheck/HRMinutesIBM p570Closest competition62,59414443,689206Total2238Banking3832Payslip84136PayrollPaycheck/HRMinutesPaycheck/HRMinutesIBM p570Closest competition43% faster! Figure 8. Global payroll France VIO test results VIO Server Tests using HR Self Service Tests using HR Self Service showed little difference with or without using VIO. This was an OLTP benchmark and not a heavy I/O user. Results are shown in figure 9 below. Virtual IO Server Tests Virtual IO Server Tests HR Self ServiceHR Self Service85%/1%CPU/IOWait1.39Average Save1.42Average SearchBaseline QTP Results 2000 LR Users85%/1%CPU/IOWait1.39Average Save1.42Average SearchBaseline QTP Results 2000 LR Users83%/1%CPU/IOWait1.33Average Save1.39Average Search2000 LR Users - VIO DS480083%/1%CPU/IOWait1.33Average Save1.39Average Search2000 LR Users - VIO DS4800HR Self Service online average response times showed improvement using VIO server vs baseline with direct connect disks in this testCPU utilization shows how busy the system was during the test runHR CPU Utilization0204060801001201 36 71 106 141 176 211 246 281 316 351 386 421%sys CPU%User CPU Figure 9. HR self service VIO test results IBM Advanced POWER Virtualization and Oracle s PeopleSoft Enterprise Version 11/06/2006 http://www.ibm.com/support/techdocs Copyright 2006, IBM Corporation Page 12 Untitled Document IBM Advanced POWER Virtualization and Oracle s PeopleSoft Enterprise Version 11/06/2006 http://www.ibm.com/support/techdocsVirtual Ethernet A complex aspect of the virtual Ethernet option is that there are two different types of Ethernet interfaces: " Virtual Ethernet - This technology enables IP-based communication between logical partitions on the same system using a VLAN and is different from shared Ethernet in that there is no connec-tion to a physical Ethernet adapter capable secure software switch (POWER Hypervisor) in POWER5 system. Virtual Ethernet adapter which connects to a physical Ethernet network. The ability to securely share Ethernet bandwidth across multiple partitions increases hardware utiliza-tion. Virtual Ethernet can be used to connect to a physical Ethernet adapter which connects to a physical Ethernet network by implementing shared Ethernet adapter. " Shared Ethernet The Shared Ethernet adapter acts as an OSI Layer 2 device bridge between the virtual adapters and the physical adapters. The bridge performs the function of a MAC ad-dress relay and is independent of any higher layer protocol. When an IP host sends an IP data-gram to another host on a network connected by a bridge, it sends the datagram directly to the host and the datagram crosses the bridge without the sending IP host being aware of it, allowing better performance utilization. Virtual AdapterDevice Driver Virtual AdapterVirtual LANLayer 2 Bridge (Shared Ethernet Adapter Physical Adapter Device Driver Device Driver Client PartitionsExternal LAN/WAN Figure 10. Virtual Ethernet LAN Copyright 2006, IBM Corporation Page 13 Untitled Document IBM Advanced POWER Virtualization and Oracle s PeopleSoft Enterprise Version 11/06/2006 http://www.ibm.com/support/techdocsVirtual Ethernet Performance These tests showed comparable performance using virtual Ethernet vs. direct connect Ethernet. The tests even showed some increase in performance with VLAN. See figure 11 for the test results. There are two aspects to tuning Ethernet performance: " The first is regular TCP/IP tuning (e.g. tuning MTU size), which is identical to what would be done on a physical network. " The second aspect, which is unique to virtual Ethernet, is the CPU capacity entitlement factor, adding bandwidth to the network . The ability to securely share Ethernet bandwidth across mul-tiple partitions increases hardware utilization. Virtual IO Server Ethernet Tests Virtual IO Server Ethernet Tests HR Self ServiceHR Self ServiceShared LAN adapters showed performance improvement in this test. CPU utilization shows how busy the system was during the test run85%/1%CPU/IOWait1.39Average Save1.42Average SearchBaseline QTP Results 2000 LR Users85%/1%CPU/IOWait1.39Average Save1.42Average SearchBaseline QTP Results 2000 LR Users83%/0%CPU/IOWait1.35Average Save1.38Average Search2000 LR Users - VIO for Ethernet83%/0%CPU/IOWait1.35Average Save1.38Average Search2000 LR Users - VIO for EthernetHR Online CPU utilization0204060801001201 36 71 106 141 176 211 246 281 316 351 386 421%Sys CPU%User CPU Figure 11. VIO Server - Virtual Ethernet test using PeopleSoft HR online self-service 8.9 Conclusions Virtual CPU Scenario 1 = 2 LPARs HR on 1 LPAR with higher priority virtual partitions and Global Payroll on 1 LPAR " Improved performance over baseline runs " Maintained Global Payroll run-time Virtual CPU Scenario 2 = 2 LPARs Global Payroll on 1 LPAR with higher priority virtual partitions and HR on 1 LPAR " Improved performance over baseline runs " Maintained HR user response time VIO Testing Benchmark Conclusion: " HR Online using VIO server for the database. Response times same as dedicated IO devices. " Global payroll France demonstrated comparable performance with VIO vs. dedicated IO " Virtual Ethernet increased performance and network capacity over baseline runs Copyright 2006, IBM Corporation Page 14 Untitled Document IBM Advanced POWER Virtualization and Oracle s PeopleSoft Enterprise Version 11/06/2006 http://www.ibm.com/support/techdocsAppendix A System p" and Virtualization Using IBM's Advanced POWER Virtualization features can help simplify and optimize your IT infrastruc-ture. Available on most IBM System p5 as optional or standard this set of comprehensive systems technologies and services are designed to enable you to aggregate and manage resources via a consoli-dated, logical view. Key Benefits of deploying System p Virtualization: " Can help lower the cost of existing infrastructure by up to 62% " Can increase business flexibility allowing you to meet anticipated and unanticipated needs " Can reduce the complexity to grow your infrastructure " Leverage 39 years of leadership in Virtualization from IBM Supported System p models and more information about APV can be found at the websites: ibm.com/systems/pibm.com/systems/p/apv Copyright 2006, IBM Corporation Page 15 Untitled Document IBM Advanced POWER Virtualization and Oracle s PeopleSoft Enterprise Version 11/06/2006 http://www.ibm.com/support/techdocsReferences IBM Redbook - APV on IBM System p5 sg247940 For more information For additional information on integrated, collaborative enterprise solutions from IBM and PeopleSoft Enterprise, call 1 888 426-5505 or visit ibm.com/solutions/oracle. Copyright 2006, IBM Corporation Page 16 Untitled Document Copyright IBM Corporation 2006 IBM Corporation Integrated Marketing Communications Server Group Route 100 Somers, NY 10589 U.S.A. ibm.com Visit www.ibm.com/pc/safecomputing periodically for the latest information on safe and effective computing. Warranty Information: For a copy of applicable product warranties, write to: Warranty Information, P.O. Box 12195, RTP, NC 27709, Attn: Dept. JDJA/B203. IBM makes no representation or warranty regarding third-party products or services including those designated as ServerProven or ClusterProven. This publication was developed for products and services offered in the United States. IBM may not offer the products, services or features discussed in this document in other countries. Information is subject to change without notice. Consult your local IBM representative for information on offerings available in your area. The examples given in this paper are hypothetical examples of how a customer can use the products described herein and examples of potential cost or efficiency savings are not based on any actual case study. There is no guarantee of comparable results. Many factors determine the sizing requirements and performance of a systems architecture. IBM assumes no liability for the methodology used for determining the configurations recommended in this document nor for the results it provides. Any performance data contained in this presentation was determined in a controlled environment. Therefore, the results obtained in other operating environments may vary significantly. Some measurements quoted in this presentation may have been made on development-level systems. There is no guarantee these measurements will be the same on generally-available systems. Some measurements quoted in this presentation may have been estimated through extrapolation. Actual results may vary. Users of this presentation should verify the applicable data for their specific environment. Information in this presentation concerning non-IBM products was obtained from the suppliers of these products, published announcement material or other publicly available sources. IBM has not tested these products and cannot confirm the accuracy of performance, compatibility or any other claims related to non-IBM products. Questions on the capabilities of non-IBM products should be addressed to the suppliers of those products. IBM, the IBM logo, AIX 5L, i5/OS, micro-partitioning, POWER, POWER Hypervisor, POWER5, System p, System p5, and System Storage are trademarks or registered trademarks of IBM Corporation in theUnited States, other countries, or both. Linux is a trademark of Linus Torvalds in the United States, other countries, or both. Oracle, JD Edwards, PeopleSoft, and Siebel are registered trademarks of Oracle Corporation and/or its affiliates. Other company, product, or service names may be trademarks or service marks of others.
