An empirical study of reliability and cost in building dependable software
Peggy Otsubo · 1992
The Back-to-Back testing paradigm is a promising technique to expedite the software testing process and increase the reliability of software. It consists of developing two or more functionally equivalent software versions, each developed independently from different formal specifications. The reliability and cost of Back-to-Back testing is empirically studied in the Multi-Version Software Phase Three Experiment. Software reliability is estimated using the Musa Software Reliability Growth Model. The results are statistically analyzed to determined if testing group size or specification language significantly affects software reliability. We have discovered that group size does indeed significantly affect software reliability, with multi-program testing groups producing more reliable software than single-program testing groups. However, three-person testing groups do not produce significantly more reliable software than two-person testing groups. For systems with low reliability requirements, single-program testing groups are the least costly and reach higher reliability levels most quickly. However, for systems with higher reliability requirements, multi-program development may be required.