DYNAMICDATA-AWARESCHEDULINGFORGRIDENVIRONMENTSUSINGADISTRIBUTEDMULTI-AGENTAPPROACHPh.D.PROPOSALDEFENSE.docVIP

DYNAMIC DATA-AWARE SCHEDULING FOR GRID ENVIRONMENTS USING A DISTRIBUTED MULTI-AGENT APPROACH Ph.D. PROPOSAL DEFENSE By S.ANGELA MERCY Supervisor: Dr NORALIZA Co-Supervisor: Dr NOOR RAIHAN ABDUL HAMID 1. Motivation and Objectives 1.1 Introduction Grid Computing has emerged as a challenging technology in the dome of distributed computing, that provides a framework for exploiting the underutilized resources in an organization and extorts enormous power in terms of disk, memory space, CPU cycles etc. and thereby increasing the efficiency of resource utilization. Grid Computing offers the highest productivity, utilizing the existing infrastructure, because of the increased computational capacity offered by the Grid. One of the most challenging issues from a Grid Infrastructure standpoint is the efficient scheduling of jobs on distributed resources. A system that maps a job to a resource is called the scheduler. Scheduling jobs in a Grid Computing environment is an intricate problem, as it involves the usage of heterogeneous resources that are geographically distributed having different usage policies, highly non-uniform performance characteristics and that have varying loads and availability. The target of a high-performance scheduler is to optimize performance in such an environment. 1.2 Issue In today’s world, increasing number of applications deal with massive amount of data (Data-Intensive applications). Examples of such applications are bioinformatics, data mining, image processing etc. Such applications tend to be deployed on grids that provide powerful processing capabilities at reasonable cost. One fundamental scheduling issue that arises when exploiting grids with these types of applications is the data management. Data Management in Grid Computing addresses the need to eliminate data access bottlenecks, by minimizing the data transfer done while executing the tasks [10] [11]. There are situations where it becomes too expensive to move data within a grid in co