Software Reliability Modelling

Duration: 1985-1988 and 1986-90
Funding source: U.K. Alvey project
Project Code: ALV/PRJ/SE/072
Partners 1985-1989:City University only
Partners 1986-1990: City University (CSR) with GEC, British Aerospace, National Centre of Systems Reliability, Logica, GEC Software, Trent Polytechnic, ICL, University of Newcastle upon Tyne
Funding to CSR 1986-1990 £750,000
Contact: Peter Mellor (p.mellor@csr.city.ac.uk)

Synopsis: As the title suggests, the objective of this project was to advance our ability to model and measure the reliability of software. An important part of this was an attempt to improve the ability of the then current software reliability growth models to give measures and predictions of reliability that could be trusted by a user. A serious problem with the reliability models was that although there are many of these in the literature, a potential user cannot trust a particular one to be better than others in giving accurate reliability predictions. Indeed, it is not even possible to match a model to a source of failure data a priori in order to be confident of the accuracy of the predictions. A major achievement was our development of several diagnostic techniques for predictive accuracy, which allow a user to fit many models to a particular data source and select the one (or more) that is giving accurate results. It is now generally possible, therefore, to obtain reliability predictions and know that they are accurate.