Nonhomogeneous Error Detection Rate Models for Software Reliability Growth

Abstract

Software reliability growth models with nonhomogeneous error detection rate are discussed. The models provide a plausible description in case that the software reliability growth is influenced by two types of errors: Types 1 and 2 errors. Type 1 (Type 2) errors are easy (difficult) to be detected. The software reliability growth models considered here are described by a nonhomogeneous Poisson process. First, the software reliability growth model which uses the testing time as the unit of error detection period Is discussed. The error detection rate per error for such an error detection process Is a time-dependent function. The meaningful assessment measures for software reliability are presented. The model parameters are estimated by a method of maximum likelihood. The asymptotic properties of the maximum likelihood estimates of the model parameters are derived and their application to the estimation of the assessment measures for software reliability is obtained. The optimum release policies for operational use of a software system based on the model are presented. Finally, the discrete software reliability growth modeling with two types of errors is discussed. The number of test runs instead of the testing time is used as the unit of the error detection period.KeywordsError DetectionSoftware ReliabilitySoftware ErrorReliability GrowthSoftware Reliability Growth ModelThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.