Improvement Of Software Reliability And Evaluate Method

Software test is the important method to insure the software quality. Software reliability is the important feature of software quality. There are strong and natural relations between them: software test improve the software reliability, in turn the accurate evaluation on software reliability can provide reasonable stopping signal of software test. While reliable of hardware steadily improving, the quality insurance of software is comparatively difficult. The study of software reliability is far more from perfect compared with the study of hardware reliability. There are many works need be done.In this article, the theory and method of software test are first introduced. The importance of software error is emphasized, software testing model is reviewed. Then software reliability theory is introduced. Several important software reliability models are presented. After that, hardware reliability study is analyzed to find out the factors that lead to its success. Based on this knowledge, software reliability study is reviewed and four difficulties of software reliable problem are pointed out. This paper proposes two methods to resolve two of them. These methods are applied to the improvement of software reliable model. The result model is verified through applying it on a Coal Transportation Management System.The major works are list bellow:(1) The special difficulties of software reliable insurance are analyzed: the uncertainty of problem domain of software development; the variety of mechanisms of software implementing and the man-made nature of software development; the lacking of standards on software devise; the blur understand on the random nature of software reliability study.(2) Proposed the classification method on software errors according to whether they are random. For random software errors according to whether they are random. For random software errors, a practical method is given to distinguish whether two errors are independent.(3) Using the knowledge about the relations between random software errors, a improvement on software reliable model is made. Take the J-M model,G-O model and Nelson model as an example, an experiment about software reliability evaluation is carried on a software system called Coal Transportation Management System. The experiment result verified that improvement is valid. G-O model is more suitable to this method.