A Method for Obtaining Software Reliability Measures During Development

Reliability is a major concern for developers of large software. A number of models have been formulated to describe the failure process of software. However, because of data collection requirements these models are often not usable by the software developer until a considerable amount of integration testing has been conducted. Ideally we would like to use a reliability model early in the development process to aid the software manager. An approach is described here which allows the use of software reliability models throughout the development process. The approach employs an information theoretic measure to combine the initial assessments of the software development manager with past development data. The use of this approach with exemplary models is described.