Minimizing Message Exchanges in Agent Based Cloud Service Composition

 Abstract—Previous work demonstrated that by adopting a semi-recursive contract net protocol (SR-CNP) equipped with service capability tables (SCTs) for dynamically selecting recorded cloud agents, their services and states, Cloud agents can effectively integrate disparate Cloud resources into a unified Cloud service. However, the choice of SCT may result in large overheads with Cloud agents exchanging a considerably large number of messages to achieve high success rates in service composition. In this paper, a comprehensive set of mathematical analyses of message exchanges by Cloud agents (Broker agents, Consumer agents and Service provider agents) in cloud service composition are presented. Experiments were performed where cloud agents adopt Particle Swarm Optimization for evolving the best service composition outcomes with the aim of minimizing the average number of messages propagated while successfully composing cloud services using SR-CNP and SCTs. Empirical results obtained from an agent-based testbed reveal that agents successfully minimized the number of messages exchanged during cloud service composition.