An Efficient Fault-tolerance Technique Using Check-pointing and Replication in Grids Using Data Logs

Grid computing systems are increasingly growing importance in the present world with advances in the network technology. Grids are composed of many geographically disturbed resources, each having its own administration domain. Grid computing involves decentralized heterogeneous, geographically distributed resources that can work on a job together. Since the resource availability is dynamic in nature, the grid infrastructure is prone to failure of job lose or delay. So in order to adapt to the failure, fault tolerant mechanism must be implemented. Commonly used techniques for fault tolerance are checkpointing and load replication or job migration mechanism. To have an efficient fault tolerance mechanism this paper introduces a novel checkpointing algorithm based on real-time work load logs that reduces overhead caused due to checkpointing. The proposed system uses Job replication to ensure completion of work, optimal replication and dynamic Load balancing is used to avoid overload in any resources and to achieve maximum resource utilization and maximize output of the grid. Keywords— Grid Computing, Fault Tolerance, Work load logs