The cost of testing software

A simple quantitative cost model that allows a manager to determine the optimum amount of time needed to test a software system has been developed. At the minimum cost point, the cost of an infinitesimal increment of testing equals the marginal benefit to be gained from the increment. The model can also be used to estimate the cost of testing software if the release data is determined by some other criteria, such as reliability. Since the cost of testing software depends on the pattern of failures, the model is based on underlying software reliability models. The cost model is explicitly developed so as to apply to a wide class of reliability models. It is noted that special cases result from choosing specific models, for example, the Goel-Okumoto nonhomogeneous Poisson process model, in which all bugs have an equal impact on the failure rate, and the Musa-Okumoto logarithmic Poisson process model, in which bugs found earlier have a larger impact