Load balancing in grid computing using AI techniques

Abstract

Work load and resource management are two important factors that have to manage across the grid environment. To increase the overall efficiency of grid based infrastructure the work load across the grid environment has to manage. Hence the work load must be evenly scheduled across the grid nodes so that grid resources can be properly exploited. The technique that we have investigated in this paper is based upon the combination of genetic algorithms which is an evolutionary algorithm and artificial neural networks. Both of these techniques are applied for local grid load balancing. Genetic algorithm selects the optimal set of jobs for assigning to the grid nodes which overall minimizes the total execution time. Afterwards when optimal set of jobs is selected they are assigned to artificial neural network which selects the minimum loaded grid processor for further processing of this optimal set of jobs. We compare our proposed technique with the already existing strategies for load balancing like random algorithm, round robin algorithm, decreasing time algorithm and least connection algorithm. Results shows that our strategy gives optimal results in terms of overall time efficiency. So we can overall conclude that GA's and ANN's increase overall efficiency of job scheduling especially in case where the tasks coming for scheduling and processing nodes are continuously increasing. © 2012 Springer Science+Business Media.