Research On Reliability Modeling And Optimization Of Software System

Along with the arrival of information age, the computer technology has the unprecedented development. At the same time, the requirement for software quality is also getting higher and higher. Software reliability, whose quantitative assessment and prediction have become the focus of attention and study, is considered as an important characteristic for measuring the software quality. The software quality, especially the software reliability has already become an indispensable question, but the present production situation of software is not satisfying. An analysis on software reliability model has been made on the current situation, the problems of the previous methods have been shown in this paper, and researches have been made on approaches to software reliability modeling and optimization.Firstly, aiming at the software aging phenomenon,the problem of software reliability modeling is studied from a new prospective. A module level software rejuvenation strategy based on the threshold value is put forward. The software rejuvenation is applied in the module subsystem, and the module level software rejuvenation is used for the software reliability modeling. The predicted value is obtained from the neural network model and the stochastic Petri net based on the practical statistical data. The rejuvenation time is decided by comparing with the given threshold, based on which we can determine the optimal rejuvenation cycle.Secondly, the problem of software reliability optimization based on the genetic algorithm is explored. The concept of niche is introduced based on the problems for the low efficiency, low convergence speed of genetic algorithm in searching the global optimum and only partial optimal solutions obtained from the multimodal function. It is based on the stochastic tournament selection operator, the uniform schema crossover operator and the non-uniform mutation that the improved niche genetic algorithm for dealing with the reliability allocation issue is proposed, which can improve the search performance of the genetic algorithm.Finally, the non-markovian stochastic Petri-nets model based on the module level rejuvenation technique and the multi-module software reliability optimization model based on niche genetic algorithm are constructed respectively for the modeling and optimization of the software reliability. At the same time, the simulation experiment and the model evaluation are performed by using the mathematical statistical methods.