FAULT TOLERANCE TASK EXECUTION THROUGH COOPERATIVE COMPUTING IN GRID

Abstract

To achieve fault tolerant task execution in grid, cooperative computing system (CCS) is proposed in this paper. Grid resources with similar statistical characteristics constitute the computing nodes in CCS. CCS executes the allocated task, considered as its primary task, by organizing the computing nodes as active and active-standby. At a moment of time, one of the node acts as active node to execute the task whereas rest of the nodes act as active-standby to provide execution backup to the task. Computing nodes in CCS may fail during task execution due to the failure/exit of their corresponding resources. To maintain the fault tolerant ability of CCS, a failed node is repaired dynamically by replacing its corresponding resource with alternative matching resource from grid. The number of computing nodes in CCS is decided by optimizing the service reliability with respect to the execution overhead of the primary task. Resource usage is optimized in CCS by overloading the primary task at each active-standby node with a low priority secondary task. Active-standby nodes execute their low priority secondary tasks concurrently to providing execution backup to the primary task. Service reliability, system throughput and task delay is observed in the simulation experiments to explore the fault tolerant ability of CCS. A task set of 500 grid tasks is repeatedly executed by varying task duration and rate of resource failure. Simulation results show that CCS outperforms the existing fault tolerant approaches being used in grid. In CCS, fault tolerant task execution is achieved without compromising on account of resource utilization as well.