Analysis of necessary condition for minimal software reliability demonstration test suite

Existing software reliability demonstration testing methods mostly draw lessons from the same methods as hardware, such as probability ratio sequential testing (PRST), TRW software reliability theory, life testing method and so on. These methods base on the principle of hypothesis testing method with a specified degree of producer's risk and consumer's risk or statistical confidence, determine parameters by means of criteria of testing ending and passing, and give an estimation value of software reliability via information during testing as well. Among these methods, test cases for demonstration testing are all assumed to be selected randomly from the actual operational profile, i.e. the choosing method of test cases represents the actual using way of the software. By means of analysis, the assumption of the principle does not hold all the time, accordingly, the reliability estimation value departs far from the real value. A necessary condition for minimal software reliability demonstration test suite is proposed in this paper, the recommended value is presented, which guarantees satisfying both the first and second kinds of risk. Finally, the existing method of software reliability demonstration testing is improved.