Robust Multi-Response Based Software Reliability Modeling

Abstract

In quality engineering, robustness indicates reduction in variability in responses in an environment where some of the input parameters are difficult to control. When these variations are minimized the product will be called robust product. This chapter deals with developing a framework of Robust Multi-response based software reliability modeling and later illustrates this methodology on a large scale communication software system. The two responses under study are Time-between-detection of defects; and effort which is defined as time period between defect detection and defect elimination. The inputs considered are severity, change request priority (CR Priority) and software development phase. S/N ratios are computed on these responses. Desirability values are computed on these S/N ratios and converted into single response called overall desirability value. Optimization using response surface methodology is performed on overall desirability value and a curve fitting based reliability model is developed. Reliability is defined on the basis of Time-between-detection of defects and Time-between-detection-and-elimination of defects. The approach will help the software engineers working on software development projects, in understanding the impact of large number of input variables on more responses simultaneously. The approach will also lead to a robust software product and a reliability model relating input variables and responses.KeywordsSoftware defectsReliability modellingDesirability functionMulti-response optimizationResponse surface methodologyEffortSignal-to-Noise ratio