Evaluation of self-adaptation using invocation patterns in service oriented architectures

Abstract: The execution of service oriented applications is based on the invocation of services that perform tasks for the application. Often, multiple service providers provide the same service, with the main differentiating factor being their Quality of Service. Moreover, the quality of a service provider varies over time as it is influenced by environmental factors such as service load, software crashes, and changes in pricing. Therefore, obtaining consistent good quality for each of the required services is not an easy task. Service invocation patterns had been proposed to aid service oriented applications in this task. We can use these invocation patterns to help the application decide which service it should invoke. However, the evaluation of the applications behavior using the invocation patterns when the quality of the service providers changes over time is a problem that has not been addressed yet. To address this, this thesis investigates the utilization and generation of service invocation patterns. Concretely, I have implemented 1) a simulator of service oriented applications that use invocations patterns and 2) an adaptation engine that dynamically generates invocations patterns. The latter can be used to update the values in the service invocation patterns when the quality of the service providers changes. I have implemented the patterns according to existing definitions and integrated them into the application. I have also conducted series of experiments in order to examine the effect of self-adaptation using invocation patterns on the quality of the application. The results of the experiments conducted in this thesis indicate that adaptive invocation patterns have a positive effect on the quality of the application.