Structural models for software reliability prediction

Many previous software reliability prediction models by this author and others have concentrated on the bulk (macro) aspects of the program. This paper describes a newly developed micro model which is based on program structure. It is assumed that the program has been written in structured or modular form so that decomposition in to its constituent parts is simple. Further, we assume that via analysis of the program the decomposition can be related to several paths or other functional structures within the program. The model is constructed based upon the frequencies with which each of the j paths are run, (fj), the running time of each path, (tj), and the probability of error along each path, (qj). Several methods of calculating or measuring the fj, tj, and qj parameters are suggested. In fact it is possible to use one technique (historical data) to produce crude estimates at the start of the design, and refine the estimates with more accurate values as the design progresses. The paper concludes with the application of the model to a particular example: calculation of the roots of a quadratic equation, and a discussion of proposed experiments for validating the model.