Economic Model-Driven Cloud Service Composition

Abstract

This article considers cloud service composition from a decision analysis perspective. Traditional QoS-aware composition techniques usually consider the qualities available at the time of the composition because compositions are usually immediately consumed. This is fundamentally different in the cloud environment where the cloud service composition typically lasts for a relatively long period of time. The two most important drivers when composing cloud service are the long-term nature of the composition and the economic motivation for outsourcing tasks to the cloud. We propose an economic model, which we represent as a Bayesian network, to select and compose cloud services. We then leverage influence diagrams to model the cloud service composition. We further extend the traditional influence diagram problem to a hybrid one and adopt an extended Shenoy-Shafer architecture to solve such hybrid influence diagrams that include deterministic chance nodes. In addition, analytical and simulation results are presented to show the performance of the proposed composition approach.