x

Third-party components that can be tuned include:

Operating System

• Work with the Operating System Administrators to identify bottlenecks. If the system is maxed out on CPU and RAM, use the system utilities to identify which processes are taking up all the resources. Perhaps, some jobs can be run at off-peak hours or deferred, or run at a lower CPU priority.
• Check all configuration files to ensure they are in sync with the current releases of installed software. Upgrades may require updates to old configuration files that can improve performance overall on the system.
• Track usage patterns on the machines to identify problems and failures, which can be fixed easily.
• System administrators will be able to provide additional insight and guidelines to tune the operating systems.

Database Server

• Verify that the database optimizer runs as recommended to generate good data statistics. Example: Oracle ANALYZE command creates optimizer statistics to choose methods of fast access to the data. If the statistics were not generated properly, performance will be affected.
• Fix Indexes. Use EXPLAIN to see what the optimizer does with SQL statements. Does it do a full table scan? Does the table have appropriate indexes for the queries?
• Spread the database over disks and I/O channels. For example, partition tables and indexes over multiple disks.
• Database administrators will be able to provide additional insight and guidelines to tune the database servers.

Network

Network administrators will be able to identify and tune bottlenecks with the use of:

• Packet sniffers, such as tcpdump.
• Network protocol analyzers, such as ethereal.
• Utilities, such as netstat, ping, traceroute, and others.
• Network administrators will be able to provide additional insight and guidelines to tune the network.

Application Server

Auto-Reload. This feature automatically reloads the web application if a change is detected. According to Apache, this is useful during application development, but it requires significant run-time overhead. This is not recommended for use on deployed production applications.

For example, you can disable auto-reloading of the WebFOCUS web application file by adding the following syntax in ibi_apps.xml:

"reloadable="false"

For example, you can enable auto-reloading of the WebFOCUS web application file by adding the following syntax in ibi_apps.xml:

"reloadable="true"  

Pre-Compile JavaServer Pages. The JVM interprets a web application by dynamically compiling JSPs. Over time, the application will run faster as all files are pre-compiled.

Enable and Tune Thread Pool

• Initial Thread Pool size
  • Minimum number of threads in pool.
  • MaximumThread Pool size.
    • Maximum number of concurrent threads in pool.
  • Thread inactivity timeout
    • Time spent waiting for client response before being returned to pool.
  • Growable thread pool
    • Automatic increase in thread pool when maximum thread pool size is reached

Logging

• Start in non-debug mode.
• Ensure that the application server or web applications do not have logging enabled at a level of DEBUG. For example, if Tomcat is enabled with logging, make sure the log level is set to INFO in the log4j.properties file.

Considerations for Apache Tomcat

The following is specific to Tomcat, but we wanted to cover it briefly because Tomcat is commonly used.

Tomcat supports many connectors and each can be tuned differently. In production you may wish to:

• Set enableLookups to “false”.
• Set maxThreads and AcceptCount high enough to handle incoming traffic.
• Set compression options.

Remove unneeded applications to save memory and startup time. For example, Tomcat ROOT applications should be removed, but you can leave the Manager web application active.

Apache Portable Runtime (APR) Based Native Library for Tomcat

The Apache Portable Runtime (APR) library is fully documented on the Apache website. It has the following benefits:

• Pre-built for Windows, but must be compiled on other platforms.
• Makes Tomcat a better general-purpose web server.
• Will enable much better integration with other native web technologies.
• Overall, makes Java much more viable as a full-fledged web server platform, rather than simply a back-end focused technology.

Java Virtual Machine (JVM)

The JVM, which is used by the application server for deploying and executing web applications can be tuned, as follows:

JVM Server Modes

• Client. Use with short-running applications.
• Server. Use with 64-bit multi-threaded, long-running applications, such as WebFOCUS.
• Main differences between the two modes:
  • How they handle compiling Java byte code into native machine code.
  • How they handle the Java Heap.
  • Optimization and thread management.

JVM Memory

JVM memory system manages two types of memory: heap and non-heap.

Heap Memory

The heap may be of a fixed size or may be expanded, depending on the needs of an application. Start the JVM with the initial heap size equal to the maximum heap size. This removes the need for the JVM to resize the heap.

• You can specify a starting heap size (-Xms) and a maximum heap size (-Xmx). For example, a minimum heap size of 1024 MB and a maximum heap size of 1024 MB is specified as: -Xms1024m –Xmx1024m.
• Set the JVM MaxPermSize to 256m. For example, -XX:MaxPermSize=256m.
• To determine the size of your application, use the Windows Task Manager, or ps or top on UNIX, or print out the garbage collector log files when running in verbose mode. You can also use a JVM Profiler.

Garbage Collection

There are documented guidelines for JVM server applications, some of which are outlined below:

• Unless you have problems with pauses created from garbage collection, try granting as much memory as possible to the virtual machine. The default size (64MB) is often too small.
• Setting -Xms and -Xmx to the same value increases predictability by removing the most important sizing decision from the virtual machine. On the other hand, the virtual machine cannot compensate if you set the value too small.
• Use the -verbosegc option to track frequency and times of garbage collection, then tune the garbage collector.

JVM Monitoring

You can monitor the JVM using:

• JConsole, which is included with the JDK.
• jstat, which is included with the JDK.
• JVM Profiler.
• Java Management Extensions (JMX), for example, add the following parameters to your Tomcat startup script or Windows service.bat:

  set CATALINA_OPTS="-Dcom.sun.management.jmxremote \ -Dcom.sun.management.jmxremote.port=%my.jmx.port% \ -Dcom.sun.management.jmxremote.ssl=false \ -Dcom.sun.management.jmxremote.authenticate=false" 

x

The following are WebFOCUS components that can be tuned:

• Applications
• Reporting Server
• ReportCaster
• WebFOCUS Client

Applications

This version of this document is not intended to cover application tuning. However, at the minimum, ensure the following:

• Requests are generating optimized SQL.
• All RDBMS efficiencies are enforced.

Reporting Server

The Reporting Server has a built-in performance framework for core workspace functions and applications. These include the following:

Data Services. The Data Services feature of the WebFOCUS Reporting Server feature is useful and can be used to:

• Identify the source of an incoming request. For example, a Data Service can be created to identify requests that were sent from ReportCaster.
• Manage throughput to the back-end databases. Data Services has several properties that can control the number of concurrent requests sent to the database servers.
• Manage system resource utilization by the WebFOCUS Reporting Servers. Data Services has several properties that can be assigned thresholds for RAM and CPU usage.

You can:

• Create Data Services to service specific application requirements. For example, create a data service to process all ReportCaster requests.
• Configure and tune each Data Service differently to meet application requirements. For example, assume you want the maximum user concurrency rate for live requests to be ten and for batch requests to be three. You can create two Data services and set different maximum user concurrency rates for each. Another example, in the following image, shows limits being placed on the number of concurrent requests a user can run for InfoAssist, but not for ReportCaster jobs.

Data Adapters. There are a number of tuning parameters, depending on the types of queries and answer sets being generated. A subset of these parameters are listed below. For more information, see the Adapter Administration Manual.

• Enable OPTIMIZATION flags (SET OPTIMIZATION).
• Fetch data in larger blocks (SET FETCHSIZE).
• Set ISOLATION Level RU (Read Uncommitted).

Java Services. This feature is used for rendering graphical images on the Reporting Server, as well as for managing JDBC Data Adapters.

• Cluster the Java Services JSCOM agents for high availability, failover, and load management.
• Standard JVM tuning parameters can be applied to Java Services for memory management and garbage collection.

ReportCaster

ReportCaster Distribution Server. This is an independent Java application program. Tuning parameters include:

• Maximum Threads. This is the maximum number of active concurrent threads to manage ReportCaster scheduled jobs that are ready to be executed. Must be used in tandem with available system resources, like CPU, memory. This parameter has a default value of 6, out of which 3 are always reserved for internal use. The remaining three are used for the Data Server Maximum Connections/Threads parameter and can be split across one or more server nodes.

  For example, if Maximum Threads=15:

  • Three (3) threads are reserved for internal use.
  • Twelve (12) threads are available to be used by all of the configured WebFOCUS Reporting Server nodes.
• Data Server Maximum Connections/Threads. Control the maximum number of concurrent connections made to the WebFOCUS Reporting Server.

  For example, if Maximum Threads=15:

  • Three (3) threads are reserved for internal use.
  • Twelve (12) threads are available to be used by all of the configured WebFOCUS Reporting Server nodes.
  • There are two server nodes: EDASERVE and UNIXCLUS.
    • Data Server Maximum Connections/Threads for the thread EDASERVE has a value of 6.
    • Data Server Maximum Connections/Threads for the thread UNIXCLUS has a value of 3.

    This means that there are three threads left that can be used by EDASERVE or UNIXCLUS, or by a new node.

• Email ZIP. You can distribute the same report using EMAIL, with the ZIP option to minimize space allocation. You can also ZIP and distribute active reports for your users, and they can slice and dice that data without being connected.
• Standard JVM tuning parameters can be applied to the ReportCaster Distribution Server for memory management and garbage collection.

Report Library. This ReportCaster component can be used for archiving distributed output from scheduled jobs. Performance features include:

• PDF compression. For example, you can schedule a report that generates embedded graphical images in the distributed output and store it in compressed PDF format to the Report Library.