A New Fault-Tolerant Algorithm Based on Replication and Preemptive Migration in Cloud Computing

International audience<span id="ctl00_ctl00_cphMain_cphSection_lblAbstract" class="margin-bottom-10"&gtCloud computing is a promising paradigm thatprovides users higher computation advantages in terms of cost,flexibility, and availability. Nevertheless, with potentiallythousands of connected machines, faults become more frequent.Consequently, fault-tolerant load balancing becomes necessary inorder to optimize resources utilization while ensuring thereliability of the system. Common fault tolerance techniques incloud computing have been proposed in the literature. However, theysuffer from several shortcomings: some fault tolerance techniquesuse checkpoint-recovery which increases the average waiting timeand thus the mean response time. While other models rely on taskreplication which reduces the cloud's efficiency in terms ofresource utilization under variable loads. To address thesedeficiencies, an efficient and adaptive fault tolerant algorithmfor load balancing is proposed. Based on the CloudSim simulator,some series of test-bed scenarios are considered to assess thebehavior of the proposed algorithm.