Online Prediction and Improvement of Reliability for Service Oriented Systems

Reliability is an important metric for measuring the quality of software. Many methods have been proposed for online predicting and improving software reliability, but most of them have the following weakness: they are not able to predict software reliability on different time intervals and to locate the faulty components that cause the declining of the reliability either. This paper proposes a new method for online improvement of reliability of service composition. We use monitored failure data at ports of services to predict the reliabilities of service composition on different time intervals. If the predicted reliability is lower than the expected value, then we locate the faulty components that cause the declining of the reliability by using an improved spectrum-fault-localization (SFL) technique. The system can be automatically reconfigured to improve the system reliability by adding a component replica or replacing the faulty component. An Online Shop example is used to demonstrate the effectiveness of our method.