SQL Server load testing: synchronized tests versus random ones
Our QA department uses Mercury TestDirector to do load testing of our .NET web apps. The basic idea is that TestDirector runs through a set of URLs to load test an application, and it can gradually scale up the number of simultaneous sessions until the datacenter is filled with smoke. When the fire alarms go off, subtract one from the number of sessions, and that's the maximum load the application can sustain. (Okay, not really, but you get the idea.)
