By means of analysis of software's usage, it can be known that the software's actual usage is just like a stochastic process, whose series of usage events are extended according to some probability distribution. If the statistical characteristics of software reliability test suite is that of the implementation in practice when customers actually use the software product, that is to say, the test process is converge to the actual usage process. Thus the minimal test effort can be determined by measuring the approximate extent of statistical characteristics of test suite to that of the whole input domain. Based on the sampling probability, one determination method of the minimal software reliability test effort is proposed in this paper and the recommended value is presented. Compared to one existing method, this method is based on the same principle of the test suite's statistical characteristics, but it adopts one different mathematical method whose demonstration detail is given and finally the influence factors of the minimal test effort are discussed in detail.