Software Performance Analysis Based On Stochastic Petri Nets

Along with the unceasing expansion of software scale, the probability of software becoming failure increases. The software fault has become the main origin of software system unreliable. The software reliability model has the core effect to estimate software reliability. The models suggested at present which have certain application conditions and ranges are not able to adapt to the request of complex and volatile application environment mostly. So the application of software reliability model has brought forward new problems. In addition, the majority of software faults are the intermittence and stochastic. Therefore, the software reliability model is usually a stochastic process in form. It causes that time is more important than definition of the logic behavior of dynamic operation when analyzing software reliability. The stochastic Petri net is one kind of powerful system modeling and analysis tools in the domain of performance evaluation. It focuses each kind of the changes of system states as well as the relations between those changes. It not only is helpful to understand qualitatively the dynamic behavior of system, but also may calculate quantitatively each kind of performance index and provide the basis for the choice of the system structure and parameters. Therefore, this paper gives an approach based on stochastic Petri net that can evaluate component-based software reliability, and an integrated implementation schema is provided in detail. The method conforms to the thought that reliability evaluation of software is carried out in the early stage of software development as far as possible.It can be seen from the example analyzed that the software reliability model built by this method is able to describe the dynamic process of the changes of software well and considers as much as possible factors influencing software reliability. This method is beneficial to reducing the complicated degree of describing and analyzing the software reliability and can improve the precision of appraising and forecasting reliability. The analysis result reflects the transient reliability and steady state reliability of software so that it has provided advantageous means of analyzing software reliability after software system runs for certain time. Moreover, the evaluation can be used as the guidance of software development.Firstly, this paper discusses the concepts and characteristics of stochastic Petri net, software reliability technology and software architecture together with analyzing the relations between them. Secondly, from the aspects to the software architecture as well as the origin and transfer of software faults, this paper introduces the methods of software reliability modeling and analysis in detail. Then, this paper provides three kinds of solutions on the problem of state space explosion which usually occurs in the application of stochastic Petri net. Lastly, it gives a complete sum-up for the work of this paper, and then brings forth future prospect of research work.