Abstract
Recently fault tolerance is the essence for achieving more reliability, availability and robustness for cloud computing. This paper proposes a hybrid approach based on replication impact heuristic to provide a high level of fault tolerance. This hybrid approach has used reactive and proactive policies. The objective of this work was to used resources efficiently by reduce the consumption of these resources. Our proposed approach has used replication impact heuristic to give each task different replica instead of all tasks have the same replica. And the available resources haven't used at the same time. Also this approach has based on reliability assessment when re-execute the fail tasks. The CloudSim is a simulator which has used to evaluate the experiments of four scenarios. First scenario has used replication and resubmission technique without any improvement. Second scenario has used replication and resubmission technique based on reliability assessment. Third scenario used hybrid approach based on replication impact heuristic with random injection transient fault injection. Fourth scenario also has used hybrid approach based on replication impact heuristic without fault injection. According to these results, we decide that our approach increases VMS efficiency and guarantee most of tasks are passed during fault occurrence.