M-ANT: MODIFIED ANT ALGORITHM FOR TASK SCHEDULING IN GRID COMPUTING

Abstract

Load balanced task scheduling is very important problem in complex grid environment. Finding optimal schedules for such an environment is an NP-hard problem, and so heuristic approaches must be used. Ant-algorithm is a heuristic task scheduling algorithm which is distributable, scalable and fault tolerant. It uses the state prediction of the resources for scheduling which is necessary for effective utilization of resources. The current ant-algorithm schedules the task based on possibilities of resources. The problem with this algorithm is that it schedules the task to low possibility resource even if high possibility resource is idle but also, If the tasks are always scheduled to the resource with high possibility, then the burden on the resource may be increased and the jobs may be kept waiting in the queue waiting for the resource to be free. In this project, I propose an enhanced ant-algorithm for task scheduling in grid which gives better throughput with a controlled cost. The simulation results of various scheduling algorithms are also compared. The results also show that the enhanced version works better than the ant-algorithm. The inclusion of price factor into the ant-algorithm makes this new scheduling algorithm more suitable for wide use.