Task Scheduling Optimization in Global Grid by Genetic Algorithm Sobhan Esmaeili * PhD of IT- Computer Networks, University of Tehran Abstract: Grid is a distributed and parallel computational system where scheduling and load balancing are two important issues. Scheduling in a grid is a sort NP-hard problem considering the dynamism and diversity of resources, its uncertain nature and limitations defined by users for it. Therefore, to solve this problem, an uncertain and meta-heuristic algorithm is required to obtain a near-optimum response. Therefore, numerous studies have been conducted on meta-heuristic algorithms including genetic algorithm. In this study, it was attempted to propose a method to improve task scheduling based on genetic algorithm. In the proposed method, an evolved architecture with two independent global and local schedulers is employed where first the network structure is divided into zones based on the distance of resources from global scheduler. Then, by combining small sub clusters of resources in each zone, the main sub cluster is formed. In this method, genetic algorithm is used to select the main cluster and attribute tasks to it with the aim of ensuring load balance using threshold value and scheduling the main cluster. Results of simulation in a fair environment using MATLAB simulation tools showed that in addition to an 8% decrease in task completion time, a 21% decrease in distance and a 16% improvement in bandwidth used, the proposed method prevents premature convergence and a pause in local minimums and it ensures global optimality, compared to the best method (ECPSO) investigated recently in relation to task scheduling in grids. Keywords: Global Grid, Scheduling, Genetic Algorithm, Clustering, Optimization Corresponding Author: PhD of IT-Computer Networks-University of Tehran Email: Dr.Esmaeili@Hotmail.Com