One Size Does Not Fit All: A Group-Based Service Selection for Web-Based Business Processes

Abstract

Abstract-Web service failures or degradations directly cause operational inefficiencies and financial losses in Web service-based business processes (WSBP). In today's competitive Web service market a considerable number of functionallysimilar Web services offered by different providers are differentiated by competitive QoS levels and pricing structures. Consequently, dynamic and optimal Web service selection is a significant challenge to business organizations which run such WSBPs. Having a cost-effective and efficient Web service selection approach is becoming an important necessity for such organizations. Current Web service selection approaches offer "one-size-fits-all" solution, i.e., one optimal service selection for all running BP instances. However, such solutions are neither efficient nor cost-effective given that the service levels of WSBPs are associated with customers classes/profiles, e.g., Gold, Silver or Bronze. Therefore, we propose a group-based Web service selection approach,"one-size-does-not-fit-all", that optimizes multiple QoS criteria and differentiate the service selection based on the BP customers classes/profiles. Our approach shows a very good improvement over existing "onesize-fits-all" approaches; 65% average cost reduction and 30% utility value improvement.

Service-oriented Application Composition with Evolutionary Heuristics and Multiple Criteria

“…Their approach incorporates randomized reinforcement learning into the algorithm and employs a directed acyclic hypergraph representation, with 10K runs to incorporate the accumulated knowledge. The use of service cost and nine QoS measures was considered by Suleiman et al (2011) to select a web service for a credit card processing task, choosing from a set of 22 candidate web services using weighted normalized metrics. Web service selection based on a bidding approach is proposed by He et al (2014), wherein 6 web services are selected to support an automobile purchase decision, subject to response time and cost constraints.…”