Software Reliability models for the first stage of Software Projects

Abstract

A software reliability analysis for the first stage of software projects is presented. At this very first stage of testing we expect an increasing failure rate, where the usual software reliability growth models based on non homogeneous Poisson processes like the Goel-Okumoto or Musa-Okumoto can not be applied. However, our analysis involves some models that combine reliability growth with increasing failure rates like the logistic and delayed S-shaped models. Our analysis also includes a new model based on contagion as in the increasing failure rate as in the reliability growth stages. We point out that increasing failure rate stages are important to be modeled since corrective actions can be taken soon and also that this characteristics highlights under modern development methodologies which development is performed simultaneously as testing, like in Agile and TDD (Test driven development). Results of the application of those models to real datasets is shown.