Methodology for Software Reliability Prediction. Volume 1.
Final rept. Jun 83-May 87,
SCIENCE APPLICATIONS INTERNATIONAL CORP SAN DIEGO CA
Pagination or Media Count:
The variation in fault density on Air Force programs in enormous the worst programs are 390 times more error-prone than the best. Obviously, there are some critical differences in these programs that cause more errors to be introduced or left undetected. If we could solve the problem of what these differences are and how to control them, then we would have learned something fundamental about the occurrence of errors in software and how to avoid them. This report describes the results of a research and development effort to develop a methodology for predicting and estimating software reliability. A software Reliability Measurement Framework was established which spans the life cycle of a software reliability. Data from 59 systems, representing over 5 million lines of code, were analyzed and generally applicable observations about software reliability were made. A detailed approach to the collection and analysis of reliability data is also presented.
- Computer Programming and Software