Sequential Sampling Designs for Estimating Software Reliability
Metadata[+] Show full item record
For any non-trivial system, it is impossible to reach the exact reliability of software due to the complexity, cost, and time required to complete the testing. Instead, a sample of test cases can be used to estimate the overall software reliability. Our objective is to obtain the most accurate estimate of software reliability by allocating test cases among partitions. In the traditional approach, the method of allocating test cases among partitions is determined before reliability testing begins. By allocating test cases in advance, there is no opportunity to take advantage of the errors in choosing the distributions of test cases that may occur during the testing of the software. The inability to use these errors to adjust the estimate during testing is a shortcoming of a fixed sampling scheme. We applied sequential sampling schemes to make allocation decisions dynamically throughout the testing process. Under these sampling schemes, we can refine the allocation of test cases sequentially based on the information gained as the testing proceeds. Using theoretical results and Monte Carlo simulation, we have shown that the proposed sequential sampling scheme performs at least as well as the balanced sampling scheme.
Table of Contents
Introduction -- A fully sequential sampling scheme for software reliability estimation -- A Two-Stage Sampling Scheme for Software Reliability Estimation -- Summary and conclusion -- Appendix
Ph.D. (Doctor of Philosophy)