An Approach for Cost Eeectiveness Analysis of Multiversion Software Using Software Reliability Models

For a software system, there are two clear methods to achieve a target reliability. One is to increase the reliability of a given software to the desired level by spending more resources in testing, and the other is to use multiversion software to increase the reliability. In this paper we propose an approach based on software reliability models, to analyze the cost eeectiveness of multiversion software for achieving a given target reliability as compared to achieving it by using a single version system. By using software reliability models the cost of raising the reliability of a software from a given reliability to the desired reliability can be estimated as the testing time needed to achieve the target. Total time for software development can be estimated as a function of this testing time. Total estimated development time is then used as the cost metric for comparison. As an example of this approach, we have used a few of the well known software reliability models to perform the cost eeectiveness analysis. The analysis shows that a three version system is cost eeective only if the testing cost is very high and even then when the mission time for software usage is a small fraction of the mean time to failure of the software.