BACK-TO-BACK TESTING

Software testing based on program execution may consume considerable development effort. In this context the testing of two or more functionally equivalent versions of a program against each other appeals because it offers a potentially simple way of checking for the correctness of the program responses to a large number of test cases. One name for the technique is back-to-back testing. The approach can be cost-effective for new software, as well as for testing software after major modifications. The paper reviews empirical information on back-to-back testing and analyses the strategy. A simple model is used to explore its efficiency and cost-effectiveness. The key parameters are the degree and nature of the interversion failure correlation, the per failure identification and correction effort required during single-version program testing, and the shape of single and multiversion reliability growth functions. The advantages and deficiencies of the technique are discussed and some suggestions given about its practical use. © 1989.