Oracle s JD Edwards EnterpriseOne 8.11 Tuning Guidelines for the IBM eserver pseries

Transcription

1 IBM / Oracle International Competency Center September 2005 Oracle s JD Edwards EnterpriseOne 8.11 Tuning Guidelines for the IBM eserver pseries By Larry Grubbs pseries IT Specialist IBM / Oracle International Competency Center

2 Page 2 Table of contents 12 Table Introduction of contents 12 Highlights Test environment for establishing 3 Subheads recommendations 43 Bulleted JD Edwards copy EnterpriseOne optimization for and POWER4 illustrations or 5 Graphics 5 Trademark POWER5 s and special special 4 WebSphere characters application server 65 Quotations HTTP server 76 Charts JD Edwards EnterpriseOne JAS 8 Legal server copy 7 Oracle database settings Introduction The reduced total cost of ownership and an increasingly internet sophisticated enterprise user base has driven the demand of the JD Edwards EnterpriseOne HTML solution. This increased demand in turn calls for a tuning guide that will optimize the performance of such a system. For the best performance of the HTML client, parameters in the WebSphere Application Server, the HTTP Server, and the JD Edwards EnterpriseOne JAS and enterprise servers need to be tuned along with the Oracle database itself. Many of these configurable parameters are based on the number of users. Consider these parameter guidelines as a starting point. Your environment will be different from our lab and your results may vary. For example, if you run a different mix of batch jobs (UBEs), or have other applications running on your system, additional iterations may be required to optimize the performance of your implementation. Keep in mind that there are steady improvements and fixes in the IBM eserver pseries software and JD Edwards EnterpriseOne code so, over time, the guidelines established by our ongoing testing will continue to improve as well. For each configurable parameter, we have included: an explanation where practical, the actual settings for 350 users (175 users running distribution applications, 87 users running manufacturing applications, and 88 users running financial applications). These settings should be extrapolated for your environment based on your number of users. the recommendation summarized in general terms. Test environment for establishing recommendations We tested with AIX 5.3 ML02 on the pseries server with WebSphere and JD Edwards EnterpriseOne The hardware configuration that we tested was Physical 3-Tier (P3T), each tier was a separate LPAR on a single eserver pseries 570 POWER5 server. The first LPAR was set up as the JD Edwards EnterpriseOne JAS server with WebSphere, the second pseries LPAR was set up as the JD Edwards EnterpriseOne enterprise server, and the last pseries LPAR was set up as the JD Edwards EnterpriseOne Database server. All partitions were initially setup as Dedicated Processor partitions.

3 Page 3 All pseries servers used in this testing had the level of software installed as shown in the table below. This combination of software offered the latest fixes available at the time of the testing and is included for your reference. Keep in mind that software does change and improve over time, and we generally recommend that customers stay current on service packs. Check the JD Edwards Customer Connection web site for latest versions of supported software. Software for test environment Description Service Pack: AIX 5.3 ML02 Oracle Database Fixpak 6 JD Edwards EnterpriseOne System HTTP WebSphere CF3 Java Table 1. Test environment update levels JD Edwards EnterpriseOne optimization for POWER4 or POWER5 The JD Edwards EnterpriseOne Optimization Flags section of the jde.ini file was changed. In addition to the already present O2 flag which is a direct result of previous IBM testing, include the qarch=pwr5 flag, or qarch=pw4 for POWER4 architecture. This flag optimizes the code for the POWER4 or POWER5 architecture. See below for an actual copy of the BSFN Build section of the jde.ini file used. The OptimizationFlags section has been Bold printed with the new compiler flag additions being underlined. [BSFN BUILD] BuildArea=/u01/peoplesoft/b9/packages OptimizationFlags=-O2 -qarch=pwr5 (OptimizationFlags=-O2 qarch=pwr4) DebugFlags=-g -qfullpath -qdbextra -D_DEBUG -DJDEDEBUG InliningFlags= DefineFlags=-DKERNEL -DPRODUCTION_VERSION -DNATURAL_ALIGNMENT CompilerFlags=-qalign=natural -qflag=i:i -c OSReleaseLevel= LinkFlags=-bI:/u01/peoplesoft/b9/system/bin32/funclist.imp -bm:sre -bexpall -brtl -lc -lm -bnoentry -L. - L/u01/ peoplesoft/b9/system/lib -ljdelib -lcallobj -lerror -lgentext -ljdb -ljde_erk -ljdecache -ljdeddapi -ljdeknet - ljderepl -ljdeschr -ljdesec -ljdespec -ljdetam -llanguage -lmisc -lpackage -lport -lqueueknl -lruntime -lsrc -l transmon -lube -lworkflow -ljdesaw -ljdenet -lowver -ljdeunicode -lv_verify -bloadmap:loadmap LinkLibraries=v_verify.so SimultaneousBuilds=5 This change should be followed by a package build and redeployment of this package. After the package deployment the change should net up to a 3% decrease of the JD Edwards EnterpriseOne CPU.

4 Page 4 WebSphere application server Most of these settings are changed by starting the WebSphere Administrative Console using a Web browser. WebSphere Application Server instances (or JVMs) HTML Clients use a web browser running Internet Explorer 6.0 which connects through the HTTP Server running on the pseries server to pass requests to WebSphere running JD Edwards EnterpriseOne JAS servlets and JSPs. We followed the JAS installation instructions for creating the JAS server instance (referred to as a JVM because each instance runs in its own Java Virtual Machine). We found that each JVM provided the best performance if it supported up to 350 users. Running 350 or fewer users per JVM reduces the memory contention within WebSphere and therefore reduces the response time for the HTML clients. If you have more than 350 users, you will need to create additional instances to provide for the additional users, as needed. 350 user example: we created one JVM to accommodate our 350 users. Recommendation: Follow the JAS installation instructions and create one or more instances to run the JAS code for up to 350 users running in each instance. Heap size memory settings You can control how much of the pseries server s total memory is available for the JAS server by varying the heap size. The way you vary the heap size for each instance is via the command line arguments within each JVM. WebSphere allows for an initial heap setting and a maximum heap setting. We found that WebSphere on the pseries server performed best when both the initial and maximum heap size settings were the same. In our test, we used both 768 MB and 1024MB heap size. We found using anything higher than 1024 MB will cause excess garbage collection times, thus affecting overall response times. A 1024 MB heap size seams to be ideal for most situations. However, 768 MB would work when a shortage of physical memory is present. Note that the response times may suffer due to smaller heap size. You can review your heap size by using the JD Edwards EnterpriseOne s web saw. 350 user example: we set the JVM s command line parameters to include Xms1024m for initial and Xmx1024m for maximum. Recommendation: for the heap size in each JVM, on the command line, set an initial heap of 1024 MB memory per user and a maximum heap of 1024 MB.

5 Page 5 JAS servlet connections Within the instance, we configured the servlet running the JD Edwards EnterpriseOne JAS code to control how many concurrent servlet request that a JVM can accept. Because we allow 350 maxclients in the HTTP Server, we set the maximum number of connections for the servlet to be 350. The maximum should be equal to or less than the number of HTTP Server child process. This setting should not exceed 350 per instance because each user uses a thread and we only want to allow 350 threads per instance. Because more of a user s time is spent on data input than running the servlet, changing this setting did not significantly affect performance. 350 user example: we set the maximum number of connections for the servlet to 350. Recommendation: set the number of JD Edwards EnterpriseOne servlet connections to equal the number of users connecting to the WebSphere instance, with a maximum of 350. HTTP server All of the tuning in this section for the HTTP Server involves changing the HTTP configuration file. The following shows how to access the HTTP configuration file using VI Editor. On the pseries server command line, type in and hit Enter: cd /usr/httpserver/conf and then open the httpd.conf file with the vi editor: vi httpd.conf At this point you now have the httpd.conf file open and ready to edit. When finished with the httpd.conf file type the following to save your changes: :wq If you wish to exit without saving your changes type: :q! Server-Pool size regulation Rather than making you guess how many server processes you need, Apache dynamically adapts to the load it sees. That is, it tries to maintain enough server processes to handle the current load, plus a few spare servers to handle transient load spikes (e.g., multiple simultaneous request from a single Internet Explorer browser). It does this by periodically checking how many servers are waiting for a request. If there are fewer than MinSpareServers, it creates a new spare. If there are more than MaxSpareServers, some of the spares die off. 350 user example: MinSpareServers 5 MaxSpareServers 50

6 Page 6 Recommendation: The default settings shown above should be sufficient for 350 users. MaxClients The MaxClients HTTP server parameter limits the total number of servers running, i.e., limits the number of clients who can simultaneously connect. If this limit is ever reached, clients will be locked out, so it should not be set too low. It is intended mainly as a brake to keep a runaway server from taking AIX with it as it spirals down. 350 user example: MaxClients 500 Recommendation: Set this number equal to the total number of HTML clients connecting to the system. You must then stop and restart your HTTP Server for this change to take effect. JD Edwards EnterpriseOne JAS server All of the tuning in this section for the JD Edwards EnterpriseOne JAS Server involves changing either the jdbj.ini or the jdelog.properties file. The following shows how to access the jdbj.ini file using VI Editor, access the jdelog.properties the same way. Keep in mind that each JVM you create will have an associated jdbj.ini & jdelog.properties file, thus all of them will need to be changed if one is changed. On the pseries server command line, type in and hit Enter: cd /usr/websphere/appserver/installedapps/ nodename /EA_JS_ portname _ nodename.ear/webc lient.war/web-inf/jdbj.ini and then open the jdbj.ini file with the vi editor: vi jdbj.ini At this point you now have the jdbj.ini file open and ready to edit. When you are finished editing the file and wish to close and save the file type: :wq To edit the jdelog.properties file, just replace jdbj.ini with jdelog.properties in the above instructions as they are both located in the same directory. JDBC Connection pool settings in the jdbj.ini JDBC is the database interface used between the JD Edwards EnterpriseOne JAS code running on WebSphere and the database enterprise server. We ran with various numbers of initial, minimum, maximum, and pool growth settings. These settings affect the number of QZDASOINIT and/or QSQSRVR jobs that get used. Connection pooling allows HTML clients to use fewer of these jobs than fat and TSE clients because they can be reused. While running with 350 users, we found that, having a base amount for minimum number connections and small growth rate up to a maximum equal to the number of users,

7 Page 7 delivered the best performance. Initial connections can be set to one for every 10 users. 350 user example: [JDBj-CONNECTION POOL] MaxConnection=350 MinConnection=5 PoolGrowth=5 InitialConnection=30 (350 users/10 connections per user=30) Recommendation: if the JD Edwards EnterpriseOne userids are all mapped to a single pseries server system user profile (proxy userid), set initial connections to 1 for every 10 users, minimum connections to 5, pool growth to 5, and maximum connections equal to the number of users, up to a maximum of 400 per instance. Turn off Debug in jdelog.properties JD Edwards EnterpriseOne 8.11 uses log4j for logging in JD Edwards EnterpriseOne Web Runtime and JAS. The logging is controlled by a configuration file called jdelog.properties. Currently the setting in this file has enabled debug level logging at the root level. Since the component level settings in this file are not enabled, no logging occurs but the logging modules are called frequently. This has the potential to severely impact performance of the web solution. Fix/Workaround: If logging is disabled for the web components, change following settings in the jdelog.properties file to improve the performance. Possible Settings: jdelog.rootlogger=debug,jdelog,jaslog New Settings: jdelog.rootlogger=error,jdelog,jaslog Oracle database settings JD Edwards EnterpriseOne 8.11 running on Oracle requires a few tuning parameters to run efficiently. The first of these tuning parameters is Oracle DB Processes. By default Oracle runs 150 processes total. In a JD Edwards EnterpriseOne 8.11 environment this ration of Oracle DB processes/user should be in the range of 10/12 processes per user. So for 150 Oracle DB processes you could run approximately 15 JD Edwards EnterpriseOne users effectively with the default settings. Ideally you would take the number of intended active users to be run concurrently and divide this number by 10 to determine the appropriate number of Oracle DB processes to run.

8 Page 8 It should be changed in the following way: Logon to SQL Server, usually sqlserver / as sysdba will do the trick. Once logged in to determine the current level of set processes type: show parameter process this will show the current setting. To change the parameter to the correct value type: alter system set processes=xxxx scope=spfile;

9 Page 9 Contact information Larry Grubbs pseries Specialist IBM / Oracle International Competency Center lwgrubbs@us.ibm.com Boyd Fenton Manager IBM / Oracle International Competency Center bafento@us.ibm.com For more information For additional information on integrated, collaborative enterprise solutions from IBM and Oracle, call or visit ibm.com/solutions/businesssolutions/oracle

10 Copyright IBM Corporation 2005 IBM Corporation Integrated Marketing Communications Server Group Route 100 Somers, NY U.S.A. ibm.com Visit 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, the e-business logo, AIX, eserver, POWER4, POWER5, pseries, and WebSphere are trademarks or registered trademarks of IBM Corporation in theunited States, other countries, or both. Java and all Java-based trademarks are trademarks of Sun Microsystems, Inc. in the United States, other countries, or both. Microsoft, Windows, Windows NT, and the Windows logo are trademarks of Microsoft Corporation in the United States, other countries, or both. Intel, Intel logo, Intel Inside, Intel Inside logo, Intel Centrino, Intel Centrino logo, Celeron, Intel Xeon, Intel SpeedStep, Itanium, and Pentium are trademarks or registered trademarks of Intel Corporation or its subsidiaries in the United States and other countries. UNIX is a registered trademark of The Open Group in the United States and other countries. Linux is a trademark of Linus Torvalds in the United States, other countries, or both. Oracle, JD Edwards, and PeopleSoft are registered trademarks of Oracle Corporation and/or its affiliates. Other names may be trademarks of their respective owners. Other company, product, or service names may be trademarks or service marks of others.