One goal of reliability modeling is to develop a model that will allow better prediction of software reliability. As software developers test their software, it would help in many ways if they knew the reliability of the software, including when determining how much more testing is needed. The authors have developed a model, the binomial software reliability model based on coverage (BMBC). Their model is based on the amount of structural testing that has been achieved before the detection of the next fault. This model has tremendous potential to improve software testing processes, the accuracy of software reliability estimation, and the basic understanding of software and faults.

Researchers in software testing and software reliability modeling will find this paper very interesting, as well as a challenge to further evaluate the potential of this model. The authors have used one example of commercial software, one set of faults discovered during development, one operational profile and one set of tests randomly generated from the operational profile to show that their model can have better correlation with the actual reliability (based on Nelson’s approach to measuring software reliability) than older reliability growth models.

Software practitioners will find little practical information to apply to real-world issues in software testing and reliability. The model was only tested in a very limited fashion, and no conclusions about its generality can be made. Additionally, practical approaches to the generation of the parameters were not discussed and will be a separate area of research. Although this model was compared with different structural testing criteria, the results for the different criteria were so close that no conclusions can be drawn.