Do Software Reliability Prediction Models Meet Industrial Perceptions?

Reliability is one of the key characteristics inherent in the concept "software quality". Understanding software reliability has been vital for both the software managers and software engineers, especially while developing critical systems. In the past several decades, an extensive body of knowledge has been developed and to a certain extent practical knowledge has been gathered by applying them to large scale mission- or safety-critical projects. However, software reliability models are falling short of gaining confidence in industrial community. From the software developer's perspective, software reliability has been primarily associated with defects, which represent a major cost factor in software development. This is called developer oriented approach. As software orientation has become ubiquitous, and software intensive systems have been proliferating, software quality assurance has been embracing more customer oriented perspectives. In this context, we will need richer, realistic and more accurate measures for software reliability. A recent empirical study by the authors highlighted many factors, which the practitioners perceived to be influential in the context of reliability prediction. This paper focuses on different reliability prediction models proposed in the literature, specifically examining whether the influential factors perceived to be contributing to the reliability are addressed or not in these models. We evaluated the assumptions made in each of the considered reliability models and checked their applicability in software product development. This paper concludes with a comparative presentation of a large set of well-known reliability models with respect to the set of factors identified by the industry as influential in software reliability.