Flexible Discrete Software Reliability Growth Model for Distributed Environment Incorporating Two Types of Imperfect Debugging

Abstract

In literature we have several software reliability growth models developed to monitor the reliability growth during the testing phase of the software development. These models typically use the calendar / execution time and hence are known as continuous time SRGM. However, very little seems to have been done in the literature to develop discrete SRGM. Discrete SRGM uses test cases in computer test runs as a unit of testing. Debugging process is usually imperfect because during testing all software faults are not completely removed as they are difficult to locate or new faults might be introduced. In real software development environment, the number of failures observed need not be same as the number of errors removed. If the number of failures observed is more than the number of faults removed then we have the case of imperfect debugging. Due to the complexity of the software system and the incomplete understanding of the software requirements, specifications and structure, the testing team may not be able to remove the fault perfectly on detection of the failure and the original fault may remain or get replaced by another fault. In this paper, we discuss a discrete software reliability growth model for distributed system considering imperfect debugging that faults are not always corrected/removed when they are detected and fault generation. The proposed model assumes that the software system consists of a finite number of reused and newly developed sub-systems. The reused sub-systems do not involve the effect of severity of the faults on the software reliability growth phenomenon because they stabilize over a period of time i.e. the growth is uniform whereas, the newly developed subsystem does involve. For newly developed component, it is assumed that removal process follows logistic growth curve due to the fact that learning of removal team grows as testing progresses. The fault removal phenomena for reused and newly developed sub-systems have been modeled separately and are summed to obtain the total fault removal phenomenon of the software system. The model has been validated on two software data sets and it is shown that the proposed model fairs comparatively better than the existing one.