Preproduction Performance Testing
Make sure that you thoroughly test your PPM system before you make it available it to users. Design your tests to exercise the most frequently performed tasks and the most frequently used portlets and filters.
Load testing is the most important preproduction testing to perform, and the best way to predict the PPM system performance.
Make sure that your load testing simulates all of the following conditions:
Multiple users accessing the PPM user nodes concurrently
Simultaneous processing of large volumes of user input, such as might occur with time-sheet entry on the last day of a pay period
Running multiple application functions based on the functional areas you plan to implement
Retrieving large data sets
Because data sets vary (in terms of both entity volumes and characteristics) across most production deployments, you cannot always predict performance.
Hardware and configuration you plan to use for production
Testing under low-user, and even single-user, load can also be a good way to predict application performance for key operations. Single-user preproduction performance testing can have minimal effect on time lines and can be incorporated into the user acceptance test cycle or other preproduction tasks.
Critical use cases are also an important part of preproduction testing. We recommend that you test actions that users perform frequently and that must perform acceptably for users to successfully complete their work. For example, if you implement Time Management, users must be able to log their time sheets efficiently with acceptable response times.
During functional or user-acceptance testing, make sure that all poor performance areas, no matter how trivial they seem, are investigated to avoid performance issues at production launch. It is always worthwhile to investigate performance concerns. Depending on the concern, you may choose to test the issue again in a staged version of your production environment or in a separate single-user load system.