The Influence of Service Interactions on Individual Service Reliability in a Composition Scenario

Selecting an optimal service from a group of functionally equivalent ones is non-trivial. This is more so when the service to be selected is part of a composite application. Research in the past has resolved this issue making use of the Quality of Service (QoS) attributes of the services to determine the most optimal from the functionally equivalents. This paper too attempts to tackle this problem using one of the more important QoS attributes, reliability. The novelty of the technique proposed here is due to the fact that while papers in the past have looked upon the reliability of individual services in a service composition in isolation, we take into account the influence that the interaction among services in a composition has on individual reliabilities. The service domain along with the interactions is represented as a continuous time Markov chain, and through appropriate procedure the reliability of individual services is calculated in the form of ‘failure distance’. The services selected are the ones with the largest values of failure distance. The results of experiments conducted by us have also been included to validate this technique.