User-centric, heuristic optimization of service composition in clouds

Abstract

With the advent of Cloud computing, there is a high potential for third-party solution providers such as composite service providers, aggregators or resellers to tie together services from different clouds to fulfill the pay-per-use demands of their customers. Customer satisfaction which is primarily based on the fulfillment of user-centric objectives is a crucial success factor to excel in such a service market. The clients' requirements, if they change over time even after the desired solution composition, may result in a failure of this approach. On the other hand, business prospects expand with the possibility of reselling already designed solutions to different customers after the underlying services become available again. The service composition strategies must cope with the above-mentioned dynamic situations. In this paper we address these challenges in context with the customer-driven service selection. We present a formal approach to map customer requirements onto functional and non-functional attributes of the services. We define a happiness measure to guarantee user satisfaction and devise a parallelizable service composition algorithm to maximize this happiness measure. We devise a heuristic approach based on historical information of service composition to rapidly react to changes in client requirements at design time and indicate run-time remedies such as for service failures. The heuristic algorithm is also useful to recompose similar solutions for different clients with matching requirements. Our algorithms are evaluated by the results of a simulation developed on the workflow tool Kepler coupled with a C++ implementation of the optimization algorithms. © 2010 Springer-Verlag.