Modeling the software failure process and software reliability

Abstract

Бог the last few deccides, the computer systems have replaced the man power in many important areas of life. Financial markets, production and service systems are controlled either fully by the computers or by the managers who extensively use the computers. Therefore, a major failure of a computerized system iriciy end up with a catastrophe that may affect a hirge number of people adversely, for this reason, reliability of computer systems attracts tlie attention of the researchers and practitioners. Software is one of the major component of a computer system. Therefore, in this thesis, we study the stochastic nature of the failure process of a software during the debugging phase. We propose a mathematical model for the software fciilure process which may also lead to prediction of future reliability. Different from many other available software relicibility models, we particularly try to incorporate the structural properties of a softwcire such as the number ol its different instruction paths cuid their logical complexities. The debugging phase is assumed to be composed of successive test sessions. Distribution functions of the length of a test session and the remaining number ol faults cifter a test session are computed. A stopping condition lor the debugging phase is proposed.