Optimizing PMF Performance

Review the following performance settings to diagnose application performance:

• DB version.
• Processor cores.
• Available memory.

Note: PMF asks the RDBMS to perform the summarization of your metrics, since transferring many millions of rows of metric data over to WebFOCUS for summarization would cause very slow performance. The SQL issued from PMF is optimized to request summaries from the RDBMS on the dimensional intersections used for the metrics.

You can check the timings for your RDBMS query by running a PMF report, or looking at the View Source in your browser, and reviewing the traces as shown below the final </html> tag.

To help determine if the RDBMS should be optimized, look for the following line:

-* ++++++++++++++++++++++++++++++++++++++++ Run time for Step(s) Extract Normal Measure Data: nn.nnn seconds

If nn.nnn is more than 3 seconds, the RDBMS should be optimized.

Note that if you set Application Tracing to ON in the PMF Settings section in the Manage tab, you will see a very detailed trace of RDBMS queries from PMF.

We recommend you:

• Use a 64-bit version of the RDBMS and OS on the RDBMS machine.
• Maximize the available memory on the box to permit in-memory processing of the core tables. Use the base row space calculation to determine the amount of memory needed. You can find this calculation in Estimating PMF Database Size.
• Check for long-running jobs running during hours when high performance is expected. Consider scheduling long-running jobs for non-peak hours.
• Use Materialized Views (Oracle or DB2) or Indexed Views (SQL Server), which provide a 10x improvement in response time for queries. These will require some planning to implement, as they must be optimized for common access. They also require processing, as they must be dropped before any changes to PMF Dimension structure are made and then rebuilt.

Review the following performance settings to diagnose application performance:

• WebFOCUS version.
• Processor cores.
• Available memory.
• Number of pre-started agents.
• Max agents.
• Is the tracing turned on?

We recommend you:

• Use a 64-bit version of both the WebFOCUS Reporting Server and the underlying OS on the WebFOCUS Server machine, if possible.
• Maximize the available memory on the box. Use the following average per-agent server load calculation to determine how much memory you will need.

  For prestarted agents, use the following calculation:

  total users * total estimated concurrency factor * total expected simultaneous demand * 8

  This determines the number of pre-started agents that you will need to support PMF concurrency with limited latency,

• Maximize available and ready agents to the maximum available processor and memory that it can handle.

  To calculate the real memory needed for agents, assume that an average memory footprint is 7MB/agent. The calculation would be pre-started agent count x 7MB. For example, if the max total users is 300, the concurrency factor is 50%, and the demand factor is 10%, you get (300 * 0.5 * 0.1) or 7.5. Multiply this by 8 agents and get 60. 60 * 7MB = 420MB of memory to support agents.

• Make sure that the maximum limit for the maximum allowed agents is identical with the number allowed by your license. If it is unlimited, use 50% above the minimum concurrency requirements.
• Ensure that WebFOCUS Server and Client tracing are disabled.
• If WebFOCUS is being used heavily for other purposes on the same box, such as running batch jobs, consider the following:
  • Scheduling long-running jobs for non-peak hours.
  • Moving any high-demand batch or non-real time jobs to another machine.
  • Configuring PMF with a dedicated running service and maximize the number of ready agents for this service. You can also increase the priority of PMF and other real-time application and decrease the priority of the batch jobs and any other job that may be running.

Review the following performance settings to diagnose application performance:

• Processor cores and available memory.
• Web server. Version, configuration.
• Is it running HTTPS or HTTP?
• Is HTTP compression turned on?
• App server (usually Tomcat). Configured heap size. Is tracing turned on?

We recommend you:

• Use a 64-bit version of the OS on the WebFOCUS Server machine.
• Maximize the available memory on the box.
• Enable HTTP Compression for the following document MIME types: HTML, HTM, XML, SVG, and JS.
• Increase available Java Heap size to the maximum that is possible for allocated memory on the box, Java Heap size alone can make a huge difference for Web App Server performance. For a 32-bit OS, the maximum is 1.5GB. For a 64-bit OS, you can double the maximum to 3GB.

Review the following performance settings to diagnose application performance:

Network. What is the speed of pipes between each choke point? Are there are special items to take note of? For example, are you logging on through a security bridge or a switch that might be slowing down traffic? How is the user conducting the test accessing the application? Are they using a T1 line, 10baseT, 100baseT, gigabit, or wifi B/G/N?

Browser. What is the browser type and version? Is the site configured to the browser as a local Intranet? Is caching turned on? Is Flex caching turned on if Flex is being tested?

We recommend you:

• Use a recent browser release. IE10 performs much faster than older versions, and allows up to 10 simultaneous HTTP threads to run on an Intranet configured application. Note that IE6, IE7, and IE8 are no longer supported by Microsoft, and is not recommended for use.
• In IE, configuring caching to Automatic permits as many files as possible to be cached. You can configure caching across your site by using a Policy. If you are running an Internet version of PMF, advise your users to turn on the browser caching feature.
• Turn on Flex caching. A number of PMF Gadgets use Adobe Flex (Flash) technology, and they are clearly marked in the Gadget catalog. If any of the Dashboards you use have Adobe Flex compounds, your users must activate and expanding their Adobe Flash caching. For more information, see the Adobe website.

PMF is a complex application with many dependencies on your RDBMS, WebFOCUS Server, WebFOCUS Client, web server, and browser. For optimal performance, and to operate PMF at full speed capacity, you need to tune your environment so it performs optimally. This is especially important if you are running the application on the Internet or in the cloud.

It is not possible for any application to speed up your network, optimize the standard web servers of your company, or increase the processing capacity of your provisioned server hardware.

Application performance is never automatic. It must be designed into the application and then enforced in the environment. Depending on the performance and response time expected, optimizing application performance might require you to review every possible choke point.

The first thing that must be determined is what is the expected performance of the application in seconds. Once you know the expected speed, you should review the entire setup of the environment in order to explore all of the possible performance choke points.

The following is a set of guidelines to maximize the response capability of your WebFOCUS environment in order to allow optimum performance of any real-time applications, including PMF.

If you need to disable minification for any reason, you can find the setting to enable or disable minification by going to the Manage tab, selecting Settings and then System. Setting Run Minimized to OFF will disable minification. The default is on and should be left ON for the best web performance.

If you system is not optimized for performance, your PMF Dashboards or other operational parts of the application might run more slowly than expected. If this situation occurs, you may see a semaphore timeout message from PMF, which indicates that AJAX processes have taken longer than expected.

You can extend the amount of time PMF waits for AJAX processes to complete by changing the AJAX Timeout setting. In the Manage tab, select Settings and then System. Set the number of seconds PMF should wait for AJAX processes to complete and return in the Ajax Timeout field. 30 seconds is the default.