A Computational Grid Scheduling Model To Minimize Turnaround Using Modified GA

This paper presents a multi entry point scheduling model for the computational grid which schedules the job to the grid resources based on the suitability of the resource and the turnaround offered to the job. The model is based on the Modified Genetic Algorithm that uses Threshold selection method. The turnaround time estimations are realistic and are being evaluated in terms of the node efficiency, speed of execution, the existing workload on the node and the job characteristics like communication between various modules of the job. Genetic Algorithm (GA) is an effective tool for the hard optimization problems. Often the result produced by the GA is sub optimal and leaves the chance of further improvement. A Modified GA (MGA) introduces an elitist selection method based on the two threshold values to improve the solution. This paper uses MGA algorithm to schedule a modular job on the grid with the objective to minimize its turnaround time. The MGA works on the basis of partitioning the population of the current generation (using threshold values) in three parts viz. the chromosomes which are the fittest, others which are not fit right now but may become fit in the coming generation and finally the ones with fitness below expectation that are often discarded. The simulation results have been compared with other grid scheduling models and it reveals the effectiveness of the proposed MGA based grid scheduling model.