Providing Fault-Tolerance in Unreliable Grid Systems Through Adaptive Checkpointing and Replication

Abstract

Abstract.As grids typically consist of autonomously managed subsystems with strongly varying resources, fault-tolerance forms an important aspect of the scheduling process of applications. Two well-known techniques for providing fault-tolerance in grids are periodic task checkpointing and replication. Both techniques mitigate the amount of work lost due to changing system availability but can introduce significant runtime overhead. The latter largely depends on the length of checkpointing interval and the chosen number of replicas, respectively. This paper presents a dynamic scheduling algorithm that switches between periodic checkpointing and replication to exploit the advantages of both techniques and to reduce the overhead. Furthermore, several novel heuristics are discussed that perform on-line adaptive tuning of the checkpointing period based on historical information on resource behavior. Simulationbased comparison of the proposed combined algorithm versus traditional strategies based on checkpointing and replication only, suggests significant reduction of average task makespan for systems with varying load.

Customizing the minimum number of replicas for achieving fault tolerance in a cloud/grid environment

“…Jiang and Zhou [17] suggested a fault-tolerant algorithm for scheduling jobs by matching the resource's trust level and the user's security, the number of copies will be determined according to the security level of the network, which is variable. Chtepen et al [18] introduced a heuristic schedule that relies on replicating functions and rearranging unsuccessful tasks using real-time network state information rather than relying on scheduled job data.…”

A job checkpointing system for computational grids

“…So, the resource failure rate (FR) is used in this paper to determine the checkpoint interval and the number of checkpoints instead of using the resource fault index. The scheduler or the broker of the present scheduling systems [5,10,13,14,16] selects resources according to the response time combined with the resource fault index to execute the job. If the selected resource is failed and it is the only available resource that can execute the job at that time, the job must wait for that resource to join the system again and become available.…”