P ERFORMANCE R EPORT O RACLE E-BUSINESS B ENCHMARK R EV. 1.1 E-BUSINESS APPLICATIONS R12 (RUP 3) RECEIVABLES ONLINE PROCESSING: Using Oracle10g on an IBM System P5 595 As a global leader in e-business applications, Oracle is committed to delivering high performance solutions that meet our customers expectations. Business software must deliver rich functionality with robust performance. This performance must be maintained at volumes that are representative of customer environments. Oracle benchmarks demonstrate our software s performance characteristics for a range of processing volumes in a specific configuration. Customers and prospects can use this information to determine the software, hardware, and network configurations necessary to support their processing volumes. The primary objective of our benchmarking effort is to provide as many data points as possible to support this important decision. TELECOM SCENARIO PROFILE In May and June 2008, Oracle conducted a benchmark in the Oracle/IBM Joint Solution Center in Montpellier (Part of the largest European support center for IBM systems) to measure the online and batch performance of Receivables flows (processes) in an environment running Oracle E- Business Suite R12 (RUP 3) with Oracle10g (10.2.0.3) RAC for the IBM AIX operating system on twin 54-CPU partitions of twin IBM System p5 595 database servers running AIX 5L V5.3 TL06 (64-bit) OS. Two (or four) 10-CPU partitions of the IBM System p5-595 servers were used as Application/Web servers. An IBM System Storage DS8300 Turbo, equipped with 384 disk arrays totaling 80 TB were used for data storage. Actual database was ~12 TB. The testing measured the Receivables batch business process hourly throughputs for a small database model. Testing was conducted in a controlled environment for OLTP and Batch flows. The goal of this Benchmark was to obtain sizing and performance results for a telecommunications scenario with 42 Million customers and 4 months of history data. SUMMARY OF OLTP RESULTS 42 Million Customers, 4 Months History Users 6,000 10,000 Transactions/Sec ~17 ~28 DB CPU (avg.) 9 cores 13 cores App. CPU (avg.) 20 cores 47 cores Many factors can influence performance and your results may differ. CONCLUSION Application throughput was measured for 6,000 users and for 10,000 users. The application was found to scale well with corresponding hardware. COPYRIGHT 2008 Oracle, Inc. All rights reserved. Page - 1
Architecture Configuration 1 Architecture Configuration 2 Two IBM p595 64-way 2.3 GHz w/256 GB Two IBM p595 64-way 2.3 GHz w/256 GB Balancing Injectors Balancing Injectors Node 1 Node 2 App. 2 94 GB 140 GB App. 1 94 GB 140 GB Node 1 Node 2 App. 2 App. 4 100 GB App. 1 App. 3 100 GB Simulated Users 6,000 10,000 Trans per Sec 17 28 Trans per Hour ~61000 ~100000 Figure 1: Summary of Test Configuration 6,000 Users 10,000 Users Transaction Users Average 90 th % Trans/sec Users Average 90 th % Trans/sec A1 A1_01_CONNECT 4358 1.577 1.311 12.36 6975 2.178 1.789 20.96 A1_02_AUTHENTICATE 0.436 0.484 12.36 1.033 1.068 20.93 A1_03_SELECT 0.574 0.635 12.36 1.579 1.33 20.70 A1_04_DISPLAY 0.542 0.574 12.35 1.346 1.18 20.66 A1_05_DISCONNECT 0.279 0.303 12.33 0.82 0.615 20.59 A2 A2_03_SELECT 1347 8.845 10.407 4.30 2156 10.96 10.675 6.47 A2_04_DISPLAY 0.544 0.615 4.30 1.595 1.288 6.43 A3a A3a_04_UNAPPLIED_TOTAL 170 88.34 96.938 0.11 275 112.948 117.605 1.24 A3a_05_QUERY_UNAPPLIED 1.083 1.299 0.12 1.629 1.423 1.23 A3a_06_UNAPPLIED_SAVE 1.626 1.813 0.11 1.998 1.969 1.24 A3b A3b_03_Forms_connect 170 3.734 3.898 0.00 275 3.715 3.807 0.00 A3b_04_UNIDENTIFIED 81.987 89.349 0.05 81.784 89.734 0.25 A3b_04_UNIDENTIFIED_TOTAL 82.193 90.07 0.05 82.216 90.116 0.25 A3b_05_QUERY_UNIDENTIFIED 0.762 0.896 0.10 0.829 0.955 0.51 A3b_06_SAVE_UNIDENTIFIED 0.883 0.986 0.10 0.932 1.032 0.50 6045 ~17 9681 ~28 Table 1: Summary of OLTP Response Times in Seconds and Realized Transactions per Second. COPYRIGHT 2008 Oracle, Inc. All rights reserved. Page - 2
BENCHMARK METHODOLOGY E-Business Suite R12 online and batch processes can be initiated from a browser. For this benchmark, all runs used a browser to initiate the on-line user transactions. HP s Runner was used as the load driver, simulating concurrent users. It submitted transactions at an average rate of one every 5 10 minutes for each concurrent user. Measurements were recorded on all of the servers when the user load was attained and the environment reached a steady state. Figure 2 shows the configuration used for this benchmark run. BENCHMARK BUSINESS PROCESSES The test included of a mix of on-line transactions and batch processes running at different times. The following table describes the on-line transactions included in the scenario. Oracle Application Product Flow % Overall Pacing in Sec Think Time A1 Self-Service Display Customer Account 72.0 320-400 15 30 A2 Support Representative Display Customer Account 22.4 320-400 15 30 A3a Accountant Update Customer Account 2.8 560-640 15 30 A3b Accountant Runner Controller 'Drivers' Runner 10,000 Users Update Suspense Account 2.8 560-640 15 30 100% Table 2: Online Processes Schedule Balancing Forms & Apache App. & Web Servers DB Server DB Server E-Bus 12-10g DS8300 Modules Figure 2: 3-Tier Configuration This benchmark was run as a Physical 3-Tier configuration with discrete machines (partitions) hosting all of the Database and Application server instances. COPYRIGHT 2008 Oracle, Inc. All rights reserved. Page - 3
SERVER PERFORMANCE Server 6,000 Phys CPUs User Sys. Wait Idle Avg 4.37 66.2 22.7 0.0 11.1 Max 7.26 72.1 38.1 0.1 21.1 Avg 4.18 65.1 24.2 0.0 10.6 Max 6.47 71.4 44.8 0.1 25.4 App. 1 Avg 10.03 78.4 18.0 0.0 3.6 Max 15.68 84.5 44.0 0.0 11.5 App. 2 Avg 10.38 83.9 12.4 0.0 3.7 10,000 Max 16.22 86.8 16.7 0.0 9.6 Avg 6.27 70.1 19.4 0.0 10.5 Max 10.24 76.7 33.1 0.1 20.7 Avg 6.66 70.2 20.5 0.0 9.3 Max 13.59 81.0 43.6 0.1 15.5 App. 1 Avg 13.16 83.7 10.7 0.0 5.6 Max 24.35 90.7 23.1 0.0 25.8 App. 2 Avg 11.47 85.7 9.2 0.0 5.1 Max 27.01 90.9 25.3 0.0 39.1 App. 3 Avg 11.52 87.4 8.6 0.0 4.0 Max 25.59 91.4 22.6 0.0 21.4 App. 4 Avg 10.57 88.2 7.1 0.0 4.7 Max 24.54 97.3 19.5 0.0 40.1 Table 3: Average Server CPU Utilization DATA COMPOSITION DESCRIPTION Major data components for the Telecommunications market model under test are summarized in the following table. Telecommunications Scenario Customers Nominal No of Customers 42,000,000 No of Person Customers 42,000,000 No of Addresses Per Customer 1 No of Contacts Per Org 1 No of Contact Points Per Customer 1 No of Contact Roles Per Contact 1 No of Site uses Per Address 1 History months 4 History Number of direct debits 101,080,000 Number of cheques 14,905,800 Number of invoices 115,985,800 Embeddded Invoices Number of unpaid invoices (to be processed by direct debits) Number of unpaid invoices (to be processed by lockbox) 2,515,800 88,200 Pickday Processing Daily inbound invoices 2,599,800 Daily inbound invoice Lines 27,455,400 Average number of lines per invoice 11 Daily outbound payments (Direct 2,515,800 debits) Daily inbound payments (Lockbox) 88,200 Invoices sent to dunning 37,800 Table 4: Data Composition COPYRIGHT 2008 Oracle, Inc. All rights reserved. Page - 4
BENCHMARK ENVIRONMENT GENERIC CONFIGURATION Servers: IBM System p5-595 and P6-595 AIX5.3/POWER5 Features: During the benchmark, three specific features of AIX/POWER systems were used: The partitioning The micro-partitioning The Simultaneous Multi-Threading (SMT) The micro-partitioning option keeps some CPUs in a general pool for both Application server and DB server to dynamically use in a shared-mode. Database: Oracle 10g (10.2.0.3) RAC SGA Size (Bytes) Database Buffers 49,056,579,584 Fixed Size 2,135,936 Redo Buffers 64,970,752 Variable Size 19,595,790,464 Network: Gigabit full duplex. On the RAC interconnect network, an ether channel has been configured, providing a maximum bandwidth of 2 gbps. The Ether channel has been built without any network switch, using a point-to-point connection between the LPARs. The links for the storage was 6 x 4Gbs fiber channel. Database Server: Two IBM System p5-595 servers were used as the database servers. The systems were equipped with the following: 54 2.3 GHz POWER5+ processor chips (SMTenabled), each with L2 Cache of 0,9 MB per core and L3 Cache of 18 MB per Core. (The Server actually has eight 8-Core modules [MCMs]. 54 cores were used for the database server and the remaining ten cores were used as a 10-core application server) Operating system: IBM AIX 5L V5.3 TL06 Storage: DS8300 75-T710 with 16 arrays of 300 GB each (Total 4.8 Terabytes.). Each array has 8 physical disks --- RAID 5, two controllers - one exclusively for redo logs. 128 GB memory cache For more details on System p5 595, please visit http://www-03.ibm.com/systems/p/hardware/highend/595/index.html For more details on DS8300, please visit http://www-03.ibm.com/systems/storage/disk/ds8000/index.html Application Servers: See above. 10 2.3 GHz POWER5+ processor chips (SMT-enabled), 36 GB assigned Application traffic management devices: 1 VLAN 81 All the LPARS (Logical Partitions) were configured with AIX 5.3 TL6 SP2 (5300-06-02-0727) using a 64-bits kernel, JFS2 system s file systems and ASM disk groups HARDWARE CONFIGURATION 1 Each p595 is configured with One 1 DB server node : 140 GB memory One APP server. 94 GB Memory DB and APP servers using SMT configuration used all the 54 CPU cores on each p595 server. Drivers: 3 IBM xseries 330 servers were used as load drivers. Each system was equipped with the following: 2 1.0 GHz Intel Pentium III processors Total Memory: 1 GB. HARDWARE CONFIGURATION 2 Each p595 is configured with One 1 DB server node : 100 GB memory Two APP servers. memory each DB and APP servers using SMT configuration used all the 54 CPU cores on each p595 server. Drivers: 8 IBM BladeCenter HS20 servers were used as load drivers. Each system was equipped with the following: 2 3.6 GHz Intel Xeon processors Total Memory: 8 GB. 6 IBM BladeCenter LS20 servers were used as load drivers. Each system was equipped with the following: COPYRIGHT 2008 Oracle, Inc. All rights reserved. Page - 5 4 2.4 GHz AMD Opteron processors Total Memory: 8 GB.
SOFTWARE VERSIONS Oracle s E-Business Suite Applications R12 (RUP 4) Oracle10g 10.X (64-bit) RAC IBM AIX 5L for POWER V5.3 with the 5300-06-02-0727 Technology Level (on the database and Application /Web/CM servers) (64-bit) Mercury Interactive's Runner 8.0 For more details on Mercury s Runner, please refer to http://www.mercury.com/us/products/performancecenter/loadrunner/ Apache WebServer 1.3.19 Java 2 Runtime Environment, Standard Edition (build 1.4.2). Classic VM (build 1.4.2, J2RE 1.4.2 IBM AIX build ca142-20061124 (SR7) (JIT enabled: jitc)) Citrix NetScaler NS7.0: Build 50.2, Date: Jan. 31, 2007 Glossary and Acronyms: RAC Real Applications Clusters Oracle Applications Performance & Benchmarks 500 Oracle Parkway Redwood Shores, California 94065 Tel 650/506-7000 Fax 650/506-7000 Email ebskit_us@oracle.com World Wide Web http://www.oracle.com E-Business Suite, AppsNet, Collaboration Suite, Oracle Direct and RAC are trademarks of Oracle, Inc. Oracle, JD Edwards, PeopleSoft and Siebel are registered trademarks of Oracle Corporation and/or its affiliates. Other names may be trademarks of their respective owners. The information contained herein is subject to change without notice. Copyright 2008 Oracle, Inc. All rights reserved. C/N 0XXX-0608 Results Registration Number: 0XXX 2008 IBM, Inc. All rights reserved. IBM, and the IBM logo are trademarks or registered trademarks of IBM, Inc. in the United States and other countries. COPYRIGHT 2008 Oracle, Inc. All rights reserved. Page - 6