Fault-Tolerant Framework for Priority-Based Service Provisioning in Cloud

Abstract

Cloud computing provides the infrastructure for posting a variety of computing applications on the back-end servers. The hosted applications may have different priorities of execution and to the satisfaction of cloud users, their applications must be executed as per their priorities. Cloud computing is inherently susceptible to failures which often result in unreliable and delayed application execution. This paper proposes a framework for priority-based fault tolerant computing service in cloud. Users' applications are characterized as CRITICAL, PREMIUIM and NORMAL in the order of their priority, with CRITICAL applications of highest priority and NORMAL applications of lowest priority. The Cooperative Computing System (CCS) proposed for grid computing is further customized in this paper for cloud computing. The customized framework is evaluated through extensive simulation experiments by executing a set of 5000 applications of mixed priorities by varying the applications' duration and resource failure rate. The promising results obtained in terms of execution delay, service reliability and system throughput for the applications of different priorities justify the proposed framework.