MODELS FOR SOFTWARE RELIABILITY.

Abstract

This paper examines several of the models proposed and used for predicting and measuring reliability of computer software. Many interesting models are only briefly mentioned and left to references. In this paper, data arising from three phases of software development are examined: design review, code production, and testing. For one software system, number of design errors detected during design review are shown to correlate well with errors for the same module discovered during pre-operational testing, leading to the hypothesis that these data may be a useful predictor of operational reliability. Some recently developed methods for assessing complexity of code productions and their correlation with errors detected in test will also be discussed.