To Improve Ant Algorithm's Grid Scheduling Using Local Search

摘要

Grid computing is a technology that represents a kind of distributed computing. This approach is mainly used in the context of resource utilization and sharing across geographical boundaries. It helps in coordinating computing capabilities, data storage and network resources across dynamically dispersed organizational structures. These areas of application inherently bring with them two major issues regarding grid computing management of resources and the scheduling of applications. The heterogeneous nature of resources in terms of architecture, power, configuration and availability complicates the problem of scheduling tasks. Reduction of makespan is the prime aim of grid scheduling. So, certain specific characteristics of the job must be considered and the metrics to be used should be decided accordingly. A very good choice for a grid scheduling environment is a heuristic algorithm called the Ant algorithm. This paper presents a proposal for a modified Ant algorithm for the Grid scheduling problem with a combination of local search. The free time of the resources and the execution time of jobs are taken into account by the proposed Ant algorithm in order to scale better utilization of resources and better scheduling of jobs. By using the standard benchmark problem, a number of intensive experiments are carried out in the evaluation study. The result illustrates that the proposed algorithm is capable of providing high quality job scheduling to grid resources. Thus the algorithm can be employed in case of real time grid environments to design efficient dynamic schedulers.