An Automatic Recovery Mechanism for Cloud Service Composition

Cloud computing, with characteristics of large scale computation, data storage, visualization, high expansibility and elasticity, provides a powerful computing paradigm. Cloud services can be rapidly composed to form on-demand composite service for accomplishing the users' requirements. However, the uncertainty of cloud services has impacted on the correctness and reliability of the composite services. Especially, for unanticipated hardware and software failures, it is very difficult to assure the quality of the composite services. In the complex cloud computing environments, recovery of the composite services from these failures is a challenging issue. The paper first presents a unified fault taxonomy in the three layers of cloud computing and analyze the causes of the faults. The authors then propose a hierarchical recovery mechanism including five different recovery algorithms for various kinds of failures. Finally, through the simulation experiments they validate the proposed approach to be effective and practical.