Imperialist competitive algorithm with PROCLUS classifier for service time optimization in cloud computing service composition

Abstract

Aiming to provide satisfying and value-added cloud composite services, suppliers put great effort into providing a large number of service providers. This goal, achieved by providing the “best” solutions, will not be guaranteed unless an efficient composite service composer is employed to choose an optimal set of required unique services (with respect to user-defined values for quality of service parameters) from the large number of provided services in the pool. Facing a wide service pool, user constraints, and a large number of required unique services in each request, the composer must solve an NP-hard problem. In this paper, CSSICA is proposed to make advances toward the lowest possible service time of composite service; in this approach, the PROCLUS classifier is used to divide cloud service providers into three categories based on total service time and assign a probability to each provider. An improved imperialist competitive algorithm is then employed to select more suitable service providers for the required unique services. Using a large real dataset, experimental and statistical studies are conducted to demonstrate that the use of clustering improved the results compared to other investigated approaches; thus, CSSICA should be considered by the composer as an efficient and scalable approach.