A New Approach to Reliability Prediction in Component-based Systems

Reliability is a qualitative feature of software that has been the focus of numerous researchers. There are several measures for quantitative evaluation of reliability, such as mean time to failure, mean time to repair, high availability, security and probability of failure of any component in the system or the whole system. In the approaches that exploit Use-case diagrams to predict reliability, the probability of failure of each use-case in the program is calculated, and using this information the reliability of whole system is achieved. However, to our knowledge those approaches consider the probability of failure of executive paths in each use-case the same. The current study intends toimprove the accuracy of pastapproach and obtain the probability of different failures for various execution paths and calculate the reliability more carefully. A web-based transaction processing system is checked to predict and evaluate the reliability. As a result, the confidence interval of the probability of failure for the entire system is .0008 which is smaller than that of the base approach.