Best practice: Running a performance test
When you run a performance test, consider the following guidelines:
- Run each test for at least 48 hours. You should run a test multiple times to get average results that accurately represent performance test in that environment.
- Use all sources of performance information. You can get useful performance information from a performance monitor, You can get useful performance information from a performance monitoring tool, the BlackBerry® Enterprise Server log files, and from the performance monitoring tools in the virtual machine software.
- Compare events as percentages and scalar numbers to even out differences. An event that happens once under one condition and twice under another condition happens 100% as often, but you must know the scalar values to decide whether the event is important.
- Look carefully at why unusual numbers occur. Some changes that performance monitors record look big, but whether the changes are big depends on the situation. For example, performance monitors often display read-write time, but the magnitude of the time depends on the placement, fragmentation, and size of a file, none of which are consistent or under your control. Some of the numbers that performance monitors display result from the type of test environment instead of the organization's environment.
Next topic: Best practice: Testing with a new workload
Previous topic: Best practice: Establishing a baseline