emp tuning培训.docx

1、emp tuning培训Email:ahmed.shahzad.awanFOUNDATION SERVER.Shared services1、 在一般情况下，采用默认的日志级别为,在调试时采用2、 设定一个组URL，虽然他是可选的也不要让他空着3、 CSS always caches groups for LDAP and MSAD4、 为确保最佳的性能，组的个数不能超过10000，如果有超过10000个组，采用一个合适的组过滤器，只保留那些你想供应的组5、 在CSS.XML的如下语句中做过滤：(|(cn=Hyp*)(cn=Marie) 6、 If client is not using gr

2、oup provisioning setMSAD - cn=LostAndFoundLDAP to any user DN node7、 设定一个用户URL，虽然他是可选的也不要让他空着8、 在Configuring User Directories时考虑更多参数进行微调，即connection pooling settings, maximum size, CSS cache refresh interval 等等Smart view1、 如果Smart view 工作区访问尽可能少的subcubes，那么它的更新会更有效率，A subcube is defined as all the r

3、ecords where Scenario, Year, Entity and Value are the same2、 你可以限制每个worksheet只有一个subcube，或者最多是两个ESSBASE ASO1、主要有以下几个地方可以优化ASO Aggregated Views Essbase.cfg数据库级别的设置MDX优化属性Aggregated Views.Perform the default aggregations as described above.Save the default selection in an aggregation script. .Turn on

4、query tracking. .Run the queries against the database, note: Queries from all query tools are tracked.After sufficient time to capture data retrieval requirements, perform another aggregation including tracked data.Analyze the proposed list of aggregate views to be stored and select the aggregate vi

5、ews that you determine provide the best balance of system resources and retrieval performance.Materialize the selected aggregate views and, saved the selection in an aggregation script.Working with aggregation scripts and various selection criteria, repeat the process until you believe you have the

6、optimum selection of aggregate views for your situation.Essbase.cfga) SERVERTHREADS 200b) AGENTTHREADS 50c) NETDELAY 1000d) NETRETRYCOUNT 1000cachese) Essbase Pending Cache Size: When millions of members, try higher values than default, e.g. 1024 Mb instead of staying with 32 Mb. Tip: Monitor the ca

7、che hit ratios (should be close to 1) in Essbase Administration Console and adjust cache size accordingly.Retrieval Caches f) 32位系统默认是10kb,64位系统默认是20kb，如果想增加Buffer Size，建议不要超过100KBg) Essbase在64位系统中需要有比32位系统更多的Sort Buffer Size ，如果你遇到一个错误说你已经超过sort buffer限制，increase the setting by a factor of two.MDX

8、Optimizationh) The NONEMPTYMEMBER property enable MDX in Essbase to query on large sets of members while skipping formula execution on non-contributing values that contain only #MISSING data.i) BSOCaches Settingsa) Index Caches最少为索引文件大小的20%尝试满足整个索引不要过度分配内存会降低性能 b) Data Caches至少要能够存储100个可扩展的数据块 可扩展的数

9、据块包含动态计算成员Data Caches要能够同时容纳200-300个数据块或者是一个父亲小的孩子成员数的最大值Data Caches越大，计算时间会越少c) Data File Caches只在Direct IO时使用最理想的是存储整个.pag文件Data Caches除以1.25Retrieval Caches j) 32位系统默认是10kb,64位系统默认是20kb，如果想增加Buffer Size，建议不要超过100KBk) Essbase在64位系统中需要有比32位系统更多的Sort Buffer Size ，如果你遇到一个错误说你已经超过sort buffer限制，increas

10、e the setting by a factor of two.Calc Cachesi. Is used by Essbase to create and track data blocks during a calc.a) Essbase.cfgb) CALCCACHEHIGH c) CALCCACHEDEFAULTd) CALCCACHELOWe) Calc script commandsf) SET CACHE HIGHii. SET CACHE DEFAULTiii. SET CACHE LOWSystem Settings:a) 设置提交块为0或者大于3000b) Essbase

11、.cfg file： SERVERTHREADS 200NETDELAY 1000NETRETRYCOUNT 1000CALCCACHE TRUECALCCACHEHIGH 50000000CALCCACHEDEFAULT 20000000CALCCACHELOW 200000CALCLOCKBLOCKHIGH 10000CALCLOCKBLOCKDEFAULT 2500CALCLOCKBLOCKLOW 100064-bit Essbase Tuning Settings considerationsa) Direct I/O 比默认的Buffered I/O访问模式更块b) 在Direct

12、I/O模式下，设置index cache 在系统资源允许范围内尽可能大(Goal: 1.00 hit ratio).c) 在Direct I/O模式下，设置data file cache 在系统资源允许范围内尽可能大(Goal: 1.00 hit ratio).d) 在Direct I/O模式下，设置data cache尽可能小(Goal: 1.00 hit ratio).e) 并行计算限制为8个进程f) 增加Retrieval Caches 到1024g) 使用不提交块方式TUNING REPORTING AND ANALYSISFinancial ReportingTune JVM Hea

13、p Sizea) Java argument for Financial Report Server SettingSuggested ValueJava ArgumentMinimum heap1024 MB-Xms1024mMaximum heap1024 MB-Xmx1024mb) Java argument for Print Server SettingSuggested ValueJava ArgumentMinimum heap256 MB-Xms256mMaximum heap512 MB-Xmx512m1、 注意：不要把Financial Management 和Report

14、ing and Analysis 部署在Windows/UNIX混合的环境下increase a number of printersa) Open file “fr_printserver.properties”:b) Default printer pool is = 5. It should be increased. Below is an explanation how to do it: c) Before following above instructions stop Reports services and remove all HRPrinter1 HRPrinter5

15、from Control Panel Printers.d) File HRCreatePrinters.ini before modificatione) File HRCreatePrinters.ini after modificationf) Run the “HRCreatePrinters.exe” fileg) Start Reports services.h) Number of printers is increased to 10:2、 If your Financial Reporting Server server has more than four cores, t

16、he value of ReportServerMaxThreads is 20. Because your server has enough resources, you can increase the value of the ReportServerMaxThreads property. Increasing the value of ReportServerMaxThreads allows the Hyperion Financial Reporting Server to execute more reports simultaneously. This property i

17、s in fr_repserver.properties. Important Note: Never set the value of ReportServerMaxThreads to 0 (zero), because this allows an infinite number of threads to be created, and memory is allocated for them even when reports are not running3、 Many variables affect report performance. Table below lists m

18、etrics by dimension to be used as a guideline for determining common terminology with regards to report size and lists metrics that define small, medium, and large reports for Financial ReportsMetricSmall ReportMedium ReportLarge ReportNumber of row dimensions11-23+Number of column dimensions11-23+N

19、umber of expanded rows100100 - 1,0001,000+Number of expanded columns1010 1515+Potential size of report1,00020,000100,000Web AnalysisJava argument for Web Analysis： SettingSuggested ValueJava ArgumentMinimum heap1200 MB-Xms1200mMaximum heap1200 MB-Xmx1200mTip: To provide additional memory for Web Ana

20、lysis servers as a 32-bit process, create additional application server instances of Web Analysis Web, each with 1.5 GB maximum heaps. When running Web Analysis as a 64-bit process on server with 24 GB RAM, Oracle recommends that you deploy five Web Analysis servers and set the maximum heap for each

21、 of the servers to 2.5 to 3.0 GB. This provides 12.5 to 15 GB of memory for Web Analysis. The maximum heap can be set larger, but large heaps require more time for garbage collection and can decrease response timesImproving Web Analysis Studio Responsivenessa) Select Start, then Control Panel, then

22、Java Plug-In to open the Java plug-in console.b) Important Note: If multiple Java Plug-in versions are installed, select the version that Web Analysis Studio uses.c) Select the Advanced tab, and add these parameters to Java Runtime Parameters:d) -Xms64m Xmx256m Xss6m (for information on parameters,

23、see the JRE documentation)To improve Web Analysis Studio performance：i. Run the SysInternal Filemon program (from ) to determine which file analyzers are using and exclude those folder and files.ii. Disable WebScanx on AnalyzerClient.jar.iii. Disable Allow Scanning Inside ArchivesInteractive Reporti

24、ngInteractive Reporting Data Access Service (DAS) Parameters修改“Relational Partial Result Cell Count”参数可以改善在关系型数据库中BQY的查询性能. 这个参数在Configuration Management Console 中的Data Access Service parameters中设置。指定DAS需要从数据库中一次性取得的字节数 ，值500,000可以很好地满足要查询时返回200,000到20,000,000个字节的数据 同样的，修改参数mdd Partial Result Cell C

25、ount可以改善在多维数据库中BQY的查询性能，对于大数据查询增大此值可以有更少的数据提取和更少的开销，但是会增加DAS的内存使用Interactive Reporting Service (BI Services) Parameters如果Interactive Reporting Service的内存使用率高，你可以修改参数DOCUMENT_UNLOAD_TIMEOUT(this setting found in server.xml for 9.3.1 and in CMC console for System 11.x)，它指出了一个thin client document在被从缓存中

26、移除之前会在缓存中保持活动状态多少时间，该值以秒为单位，默认为900秒，降低该值文件在缓存中保留时间就越短，从而释放内存也更快，但是它可能会要求用户从库中检索文件更频繁，所以在每一次安装的时候需要仔细权衡来得到最佳值如果服务器的物理内存允许，就考虑增加PROCESS_MEMORY_LIMIT值，它确定了32位的Interactive Reporting Service 什么时候可以通过进程监视器来重新启动，增加此值从默认的1.5GB到3GB可以让Interactive Reporting Service 的内存变的很大，在重新正常启动之前可以为更多的请求提供服务BI Service / DAS

27、 Processes Move or replicate the BI and DAS services on same server or to dedicated servers if more CPU power / memory are needed.Infrastructure 提示与技巧#1: 增加retrieval buffer的大小 in Administration Console, right-click the database and select Edit, and then Properties. Under Data retrieval buffers, set

28、Buffer size to 1024 注意：如果你可以使用比较多的内存，你可以配置较高的retrieval buffer 值，retrieval buffer 可以分配给每个线程，所以内存需求的数量为：线程数乘以retrieval buffer的大小#2:使用Lifecycle Management (LCM)时在IIS进程中出现“OutOfMemory” 异常 当在大型应用上运行几个Financial Management LCM 的迁移时，在IIS 进程（w3wp.exe）中可能会出现OutOfMemory异常 a) 根本原因：完成迁移后IIS没有马上释放内存，而是只能等待直到得到一个内

29、存阀值来回收内存b) 解决方法：在Financial Management Web server上修改IIS在Financial Management LCM 的application pool 上的配置i. 设置虚拟内存1000MB和物理内存800MB来启动内存回收（根据不同的硬件资源，这些值可以增加，但这些建议值对大多数环境来说应该是安全的），下面是修改pools Properties 的页面ii. 在Helalth中，修改Shutdown time limit值为10800 ，设置如此高的值是为了防止在做长时间迁移时失败#3: The first login to Workspace 1

30、1.1.1.x is slow 如果通过workspace访问的产品没有被启动过，那么从workspace登陆到产品时就会很慢a) 根本原因：这是因为workspace试图得到捆绑了每个产品的“global” 字符串时，如果产品没有启动，那么浏览器等待连接产品就会超时，第一次失败后，workspace就不会再去尝试取得该字符串了b) 解决方法：确保所有的通过workspace访问的产品已经启动，可以用EPM System Diagnostics tool 来验证所有的服务已经启动，在安装了workspace的机器是运行： Windows:- /common/validation/9.5.0.0/validate.bat UNIX:- /common/validation/9.5.0.0/validate.sh The report is generated in: /common/validation/9.5.0.0/reports The report is named: validation_report_.html 如果产品没有启动，报表会有错误显示.，否则报表就都显示为绿色, 下面是 Foundation/Workspace的例子：#4: 调整AIX 内核参数对于C+程序，如IR (BI, DAS) a