The Research Of Software Reliability Model And Software Optimal Release Problem

With the development of computer, software has become more and more important in our daily lives. For software, we consider more about the software reliability and the total cost which is related to the software release time. So it is important and practical to research software reliability and the optimal software release time. During the passed 30 years, there have been lots of software reliability models, and these models described software reliability and expense in some degree. However, there are more or less limitations in those models, and there is no one that can be used widely. So further research should be done for these models.First of all, this paper introduces mainly the problem's significance, background and development in the research field. Then software reliability's definition, correlative nomenclatures, parameter standards and elementary mathematic relations are discussed, and the basic concepts related to the optimal software release time are presented.Second, this paper elaborates on software reliability models' characteristics, sorts and composing. Some comparative research on several classical software reliability growth models from model hypothesis, elementary mathematic expressions and parameter estimates etc is done.Thirdly, this paper sets forth several classical non-homogeneous Poisson process model. Upon the basis of these models, we introduce the Pham imperfect debugging NHPP model, and describe a software reliability model considering misarrange time, which is more practical.At last, this paper analyzes and discusses the optimal software release time problem. The stopping rules of software test have been researched widely. However, there is no theorem can be used widely. This chapter first introduces the optimal software release time problem based on the Pham imperfect debugging NHPP model which is used widely now. Most of the paper discusses the minimum cost problem on condition that the reliability of the software has measured up to customers'requirements without taking the effect to the total cost caused by the reliability when released and the failure after released to the customers into account. Therefore this paper researches this problem in detail. Then the testing must be ended at some deadline. Ohterwise the developing company will pay penalty for the delay. Few paper researched this problem together with customers'required reliability and the failure of the software but this paper does. Both of the two model are examplified through a set of experieced data.