15 - Applications of time series models to software reliability analysis

Abstract

This chapter presents the applications of time series models to software reliability and illustrates the usefulness of these models for describing the behavior of software failures and for monitoring the reliability of the software. The models described in the chapter are motivated by two possible scenarios during the development stage of computer software. A Fourier series model has been introduced for describing clustered software failures. A random coefficient autoregressive process has been analyzed as a model for reliability growth. In the first scenario, the software configuration is held fixed during a test phase and modifications are incorporated as a group at the end of the phase. In such a case, the important issue is to analyze the stochastic behavior of the software failures between modifications of the software configuration. Clustering is a grouping of similar objects. For software failures, the clustering can be caused by variations in the operating environment. Software failures occur in clusters if the times between successive failures are short for some failures and long for others.