Research And Implementation Of Software Reliability Accelerated Test Technique

Software reliability is the most important factor of software quality. In highly reliable software, especial safety-critical software, the critical operation's use probability is very small. Therefore, how to guarantee: On the one hand to make the critical operation tested enough, on the other hand to assure the software reliability parameters not to emerge big deviation with the change of test models, that is the key problem of safety-critical software reliability engineering.Accelerated test is a common useful method of hardware reliability test. But its application is very seldom in software reliability engineering and test, as well not systemic. When we use accelerated test techniques in reliability engineering and test, we can find defects efficiently, accelerate the test of critical modules, decrease the cost of software development, and accelerate the release of software. In this dissertation, firstly we describe some accelerated test techniques which base on two usage models, and then we give some other accelerated test techniques from the aspect of software reliability engineering.Stratified sampling and importance sampling are two Monte-carlo statistic methods. Software Markov-chain usage model is a mathematical model and preciously describes the usage of software. In safety-critical system, not only the software reliability test often needs too much time, but also critical operations tests are often not enough and cost very highly. To resolve these problems, we describe the software reliability accelerated test which uses the method of stratified sampling and importance sampling and is based on Markov usage model. A case study shows that this method can effectively improve the test efficiency of the software reliability.The orthogonal experiment design is a scientific method which is used to search for the best (or better) experimental scheme. In software reliability test, the test cost increases rapidly with the increasing of the input numbers and operation numbers. We describe a method which is based on the orthogonal experiment design to build operational profile. To test the software reliability, this method can dispose the test input sets and then optimize the test input sets. The experiment shows that this method can simplify the building of operational profile, and decrease the test cost of software reliability.In the software practieai deveioprnent process,the serious leveis that software reriaointy factors influence the programs reflect the software reliability levels. We apply the technique of fuzzy integrated estimate into software reliability accelerated test, which can ameliorate the software reliability test, and can ameliorate and expand the software reliability models.J.D Musa indicates: when we use Run Category in software reliability test, we can enhance test efficiency. According to this thought, we give a technique to apply Run Category into software reliability test, and optimize it based on genetic K-Means algorithm.Based on the research of software reliability accelerated test techniques, this dissertation designs and implements RISA, the tool of software reliability accelerated test, which is based on the JAVA platform. As one part of the Software Reliability Engineering Tool, RISA can be used for the military software reliability accelerated test.