Software reliability using markov chain usage model

Abstract

Statistical testing gIves us opportunity to have statistical inferences such as reliability, mean time to failure (MTTF) etc. for software systems and Markov chain usage model gams it's credibility in this field. Markov chain usage model has several benefits. It allows generating test sequences from usage probability distributions, assessing statistical inferences based on analytical results associated with Markov chains and also to derive stopping criterion of the test process. But the main problem in this process is to model .software behavior in a single Markov chain. For large software systems the model size i.e. the number of sates become unwieldy and it becomes infeasible to apply this method in generating test cases as well as measuring reliability. Two Markov models called usage chain and testing chain are developed from the example software. The discriminant value of the two chains is determined to analyze software reliability. As the software becomes more complex the model size grows quickly, which is known as state explosion problem. To overcome this problem we present a technique to measure software reliability by combining the ideas drawn from stochastic modeling, statistical testing using Markov cha in usage model and component based software testing. We have taken example from database based application software, find its modules, in this case forms, and measure reliability of each forms using Markov chain usage model. We then analyze system reliability from those form's reliabilities according to their usage probabilities. Our experimental efforts lead us to a more practical and effective approach for software reliabilityand quality assurance.