What is duration testing?

While going through the skills measured section for Microsoft exam 70-518 ( http://www.microsoft.com/learning/en/us/exam.aspx?ID=70-518&locale=en-us#tab2 ) Among the various kinds of reliability testing mentioned there is "Duration Testing"

I believe there is开发者_StackOverflow中文版 no real automated way of doing this. You just need to start up your application ( probably stress testing it ... the stress can be automated though ) and observe it after a while (memory / performance). Is this correct?

I agree with your general thinking, run the application for an extended period under a realistic workload. The detail though is application-specific. For some applications there are interesting effects over time, for example some house-keeping sub-taks runs every hour, some defect in that might only manifest after many executions.

I see two major motivations for long duration tests:

1. to discover certain classes of defect that have cumulative effects and hence are most easily detected after a period of time.
2. to discover "alignment of the stars" defects - defects that only appear under some combination of circumstances - for example something odd happens when we run across midnight, or when some dependent system is taken down for maintenance at 03:00. Ideally we have a more scientific approach to finding these defects, but I do observe that duration tests do seem to discover them.

There's no such thing as duration testing. It's Endurance or more specifically "reliability testing."