Scheme For Comparing Results Of Diverse Software Versions

The paper presents a scheme for comparing the results produced by diversely designed SW versions in order to select and deliver presumably correct result. It also allows to determine all faulty versions of SW and all faulty comparators. As compared to the majority voting scheme, it requires a lesser number of result comparisons and is able, in most situations, to deliver presumably correct service even if the number of faulty SW versions is greater than the number of correct ones. The scheme is based on system-level diagnosis technique, particularly, on the comparison-based testing model. The proposed scheme can be used for designing fault-tolerant diverse servers and for improving adjudicator in N-version programming technique.