Study On Software Reliability Model And Optimal Software Release Time

In recent years, with the increasing scale of software, functionality improvementsof software system and continuously updated version of the software, the requirementsof software quality are also rising constantly. Software reliability has become a crucialstandard of software quality which becomes more and more important.This paper introduces the research significance and the general study status fromdomestic and abroad in recent years of the software reliability. Then the basic theory ofsoftware reliability, the foundation of software reliability modeling, and some classicalsoftware reliability models are introduced. Moreover, the author improves the G-Omodel and proposes a new cost model based on the improved G-O model, thenresearches the optimal software release time on the basis of the proposed cost model.The main contributions of the paper are:①The author proposes a new fault detection rate of software reliability growthmodel. In the existing model, it is generally believed that the fault detection rate isconstant or decreases as the process of software testing. This paper introduces thehuman learning ability to software reliability model, and proposes that software faultdetection rate is not only related to the number of residual faults in the software andshows a decreasing trend, but also as the human learning process to software system inthe testing period and they are constantly familiar with the software system, the faultdetection rate shows a increasing trend with the humans’ learning ability. Consideringthe two aspects conjointly, the author proposes improved fault detection rate.②Incorporating the new fault detection rate to the existing G-O model, a newsoftware reliability growth model is established. And in accordance with theassumptions of the model and the characteristics of the non-homogeneous Poissonprocess, it can be calculated the software failure rate, the expected number of softwarefaults and software reliability in the new model.③Incorporating the proposed new software reliability growth model into softwarecost model. Then improving it and building a new software cost model.④Some experiments have being done by MATLAB. The author uses geneticalgorithm to estimate the parameters in the software reliability growth model andcompares the prediction data of model with actual data. The author compares theproposed model with the existing models and proves the improved model has a better performance on the accuracy of data fitting and prediction. Then building new softwarecost model based on the improved SRGM, and it can be proved that the software costmodel exists the optimal release time. From the experiment we can derive the softwareoptimal release time and illustrate the impact of various parameters of the model to thesoftware optimal release time.