Efficient checkpoint algorithm for distributed system

Abstract

The Grid is rapidly emerging as the means for coordinated resource sharing and problem solving in multi-institutional virtual organizations while providing dependable, consistent, pervasive access to global resources. The emergence of computational Grids and the potential for seamless aggregation and interactions between distributed services and resources, has led to the start of new era of computing. Tremendously large number and the heterogeneous nature of Grid Computing resource make the resource management a significantly challenging job. Resource management scenarios often include resource discovery, resource monitoring, resource inventories, resource provisioning, fault isolation, variety of autonomic capabilities and service level management activities. Out of this fault tolerance has become the main topic of research as till date there is no single system that can be called as the complete system that will handle all the faults in grids. Checkpointing is one of the fault-tolerant techniques to restore faults and to restart job fast. The algorithms for checkpointing on distributed systems have been under study for years. These algorithms can be classified into three classes: coordinated, uncoordinated and communication-induced algorithms. In this paper, a checkpointing algorithm that has minimum checkpointing counts equivalent to periodic checkpointing algorithm has been proposed. For relatively short rollback distance at faulty situations and produces better performance rather than other algorithms in terms of task completion time, in both fault-free and faulty situations. This algorithm has been implemented in Alchemi.NET because it did not currently support any fault tolerance mechanism.