Towards Efficient Service Composition in Multi-cloud Environment

In this paper, we propose a theoretical model for the service composition in MCE. A user sends service requests to the MCE. Each service request can be satisfied from multiple clouds (i.e., service composition). Given this model, we then design a multi-layer algorithm to minimize the service composition overhead. The overhead is measured through two fundamental metrics: (1) the average number of clouds (ANC) involved in the service composition, and (2) the average number of service files (ANS) examined. While simple, the two metrics capture the fundamental communication overhead across clouds and within clouds, respectively. Preliminary evaluation based on simulation show that our algorithm outperforms the previous approaches in terms of both ANC and ANS.