Skip to Main Content
It is noted that modern complex computer-controlled systems frequently fail due to latent software design errors encountered as the software processes various input combinations during operation. Probabilistic models for such errors and their frequency of occurrence lead to software reliability functions and mean time between software error metrics. The author reviews the progress made in this field since 1970 and focuses on the successes which have been achieved with existing models. Future progress is seen as depending heavily on establishment of a database of software reliability information. This is necessary so that early, more accurate, and widespread use can be made of the proven prediction models which now exist. For creating an adequate database a number of requirements are deliniated and discussed.