An experiment in software redundancy with diverse methodologies

The goal of this experiment was to study the possibility of achieving highly reliable software using an approach to software redundancy with diverse methodologies. The experiment was similar to an experiment done by Knight and Leveson (1986) except that two different programming methodologies were used. Data from the experiment were analyzed using the simple statistical model for multiversion programming developed by Knight and Leveson and the more sophisticated statistical model of Eckhardt and Lee (1985), to see if results were consistent with the previous results for a single methodology. A still more sophisticated statistical model due to Littlewood and Miller (1989) was used to assess the effectiveness of using diverse methodologies.<>