Accurate Reliability Prediction Based on Software Structure

Software reliability growth models (SRGMs) are inadequate to assess the reliability of modern, heterogeneous, component–based software systems since these models treat the system as a black box and model its input/output behavior without looking into its internal structure. Development of techniques to assess the reliability of a component–based software system (which may be assembled from a variety of components, some picked off–the– shelf, some developed in–house and some developed contractually), based on its structure is thus absolutely essential. Most of the prior efforts in the area of structure–based reliability assessment use the composite solution approach to predict the expected reliability of the system. While the composite approach produces an accurate reliability estimate, it does not explicitly relate system reliability to the reliabilities of the individual components and system structure. The hierarchical approach, on the other hand, produces a mathematical expression relating the system reliability to the reliability of its individual components and system structure. Such an expression facilitates sensitivity analysis, exploration of alternatives through optimization and identification of reliability bottlenecks. However, the reliability estimate obtained using the hierarchical approach is only an approximation of the one obtained using the composite approach. In this paper we develop an accurate hierarchical method to estimate the reliability of a software system based on its structure. The method incorporates second–order structural statistics, and hence provides an estimate that is closer to the one produced by the composite approach. Due to the improved accuracy in the reliability estimate afforded by our method, the hierarchical approach may be used with greater confidence for other purposes. We illustrate the use of our method with a case study.