• Benchmarking with JMeter

  last modified November 2, 2007 by tesdal

We need baseline benchmark and a set of benchmarks that can be run at regular intervals to check how changes to Plone affects performance.

We use JMeter for performance tests. Test plans are stored in JMeterTestPlans for now. You can check them out with:

svn co https://svn.plone.org/svn/collective/JMeterTestPlans/trunk JMeterTestPlans

Making your own tests

Making your own tests is fairly simple. You can use bits and pieces of existing tests, but the major advantage is setting up JMeter as a proxy and click your way through the web interface, doing what you want to test.

  • In your browser settings, make sure localhost and 127.0.0.1 is not filtered in the proxy settings if you are using a local Plone instance.
  • You want to set add the plonesite variable and set it to the name of your Plone site before recording. You can see examples of this in the JMeterTestPlans files, or you can start out with one of the existing files.

Types of sites

We try to identify the most relevant types of sites

  • Plone out of the box
  • Medium sites without customizations or additional products, but lots of content and users
  • Customized sites with additional products, content types, workflows, users and content
We want GenericSetup profiles that can be applied to set up stereotype test sites. These profiles can then be applied in JMeter test plans.

Operations to profile or benchmark

  • Log in - basic test plan done
  • Content creation - basic test plan done
  • Renaming a site or folder - done
  • Browsing different parts of the site (see if cache works)
  • Workflow state change/publishing - basic test included in content creation