Determination of Software Reliability Demonstration Testing Effort Based on Importance Sampling and Prior Information

Abstract

AbstractWhen using conventional statistical methods to calculate the software reliability demonstration testing (SRDT) effort for high reliable software systems, it has a very huge number of test cases and takes a long time of testing, which leads to a fact that the conventional statistical methods cannot be put into actual usage. In this paper, according to the characteristics of the high reliable software systems, an accelerated operational profile (OP) and an acceleration factor were put forward based on importance sampling theory. For high reliable continuous-type software, by means of the prior information, a method of determination of the SRDT effort with accelerated OP and prior information was proposed by Bayesian inference method. It could significantly reduce the number of the test cases and ensure the confidence of software. Finally, the effectiveness of the proposed method was shown by a case study, and the estimation method of the hyper-parameters of prior distribution was given.Keywordsimportance samplingaccelerated operational profilesoftware reliability demonstration testingprior distributionhyper-parameters