http://dev.plone.org/collective/browser/collective.loadtesting

• Applications
• Baseline performance benchmarks for core plone development, buildbot
  
• Developers/Integrators evaluating add-on performance impacts
• Deployment configuration performance evaluation 
• Well-documented buildout
• Reproducible buildout
• Good documentation for all goals
  
• Load testing tool
• Must be able to aggregate statistics on each request type (view vs add vs edit vs workflow change)
• Ease of use and ease of extending trumps clustering
  
• JMeter
• Pros
• Wide acceptance, everyone uses it
• supports clustered load testing
• Cons
• Java
• Can only do with responses what JMeter supports
• ugly regexp stuff
  
• Funkload
• Pros
• Python
• Test plans are Python classes, easier complex test plans
• Can use Python to handle/extract from responses
  
• Cons
• doesn't support clustered load testing
• Test scenarios and content profiles
• Core buildout will have 2 test scenarios and 2 content profiles for a 2x2 grid of 4 load tests
• Test Scenarios
• Read-only public facing CMS
• Write-intensive intranet
• Content profiles
• Small: few users and groups, few folders, little bit of content
• Large: many users and groups, many folders, some large folders, lots of content, lots of binary content
• Have the buildout generate GenericSetup profiles so that they can be used for purposes other than just load testing
• Blogs
• At the Plone Performance Sprint
• Functional Benchmarking Accessibility
• Bristol Performance Sprint Post-Mortem