Efficient adaptive QoS-based service selection

Distributed service-based systems are becoming increasingly common, with a vast range of resources and functionalities being exposed as services over open networks (e.g. the web and Grid systems). Due to the distribution, participant autonomy and lack of local control, such systems operate in highly dynamic and uncertain environments, in which services can be added, removed or change their characteristics, at any time. Thus, adaptation to change during service composition is essential to meet user needs. Yet, even when service changes occur at an early stage (e.g. at selection time), current adaptive composition approaches delay their detection until after the quality violating or unavailable service is invoked, resulting in a costly recovery during execution and, in some cases, permanently unachievable goals. In response, this paper presents a novel reactive selection algorithm, which adapts to service changes efficiently while performing the selection, ensuring an executable, satisfactory and optimal solution prior to execution. The effectiveness of the algorithm is demonstrated analytically and empirically through a case study evaluation applied in the framework of learning object composition.