Parallel adaptation of multiple service composition instances

Abstract

Existing approaches for adaptation of service compositions do not consider the fact that common services can be used in different compositions, and, therefore, a problem that may be identified in one composition could be used to predict unwanted situations in other compositions. In this paper, we propose a parallel and proactive adaptation framework that supports proactive adaptation in multiple service composition instances at the same time. In the framework, events observed for one particular service composition instance are shared between all composition instances executed in parallel in order to better predict problems and rectify them in all necessary instances, when possible. The parallel characteristic of the framework also supports balancing the load among candidate service operations, and, therefore, it considers the maximum expected service operation throughput between the compositions. A prototype tool has been implemented to illustrate and evaluate the framework in different scenarios.