Prognostics and Health Management in Software Systems Engineering
Tadashi Dohi, Professor
Department of Information Engineering
Graduate School of Engineering
Hiroshima University, Japan
Prognostics and health management (PHM) is a discipline that links studies of failure mechanisms to system lifecycle management, and utilizes information to allow early detection of impending or incipient faults, remaining useful life calculations, and logical decision-making based on prediction. Since software aging and rejuvenation research can be regarded as a part of PHM in software systems engineering, the similar approach can be taken for the analysis. It is well known in statistics that the prediction of future unknown pattern should be distinguished from the estimation of past observation, and is not a trivial task under some realistic conditions. In this talk, we summarize both of traditional model-based inference techniques and machine learning techniques for the predictive inference in software reliability engineering. The most representative examples are software bug prediction which implies two meanings; prediction of fault-prone modules and quantitative prediction of the number of bugs detected in future. Another classification arises in the kind of knowledge (information) used for analysis: parametric prediction and nonparametric prediction. We concern an applicability of several techniques to software aging prediction, and introduce a few ideas to bridge between the failure phenomena and aging phenomena.