Performance Testing has long been conducted as a single experiment against a fully complete and assembled system. Architecture, software, hardware, networks, data states, and workload models are all combined to try and create the most “accurate” simulation possible to produce test results predictive of production experience.
Performance testing with this approach can be helpful for validating a completely assembled system, but in the new world of (a/A)gile development contexts, testing at the very end is unhelpful for providing timely feedback. To be more useful, performance testing can be adapted to the component levels and iteration intervals delivery teams are often working with. This video discusses approaches and techniques for providing performance feedback earlier, more specifically, and more often.
Video producer: http://www.foocafe.org/