This text comes from the old Testing portal:
Performance testing
It would likely be wise to set up some automated performance testing; e.g., checking how much time, how many DB queries, how much memory used, etc. during various operations (maybe the same tests running above).
While performance indicators can vary based on environment and non-deterministic factors, logging and graphing the results of multiple iterations from a consistent testing environment can help us with two important things:
- Identify performance regressions when we see an unexpected shift in the figures
- Confirm performance improvements when we make some optimization
See for instance Mozilla's performance test graphs for Firefox:
(more FF stuff via https://developer.mozilla.org/en/Tinderbox )