On the determination of an appropriate time for ending the software testing process

Abstract

Software testing is widely used as a means of increasing software reliability. The prohibitive nature of exhaustive testing has given rise to the problem of determining when a system has reached an acceptable reliability slate and can be released. This has probably become the hardest problem facing a project manager. In this paper, a stopping rule that indicates the appropriate time at which to stop testing is presented. The rule automatically adapts to modifications in the assumptions, since it can be applied under any software error-counting model. An investigation of the properties of the rule is described and the results obtained after applying it to a set of real data in conjunction with two statistical models are presented.