SERVICE LEVEL AGREEMENT BASED GRID SCHEDULING FOR WORKFLOW APPLICATION

Abstract

In grid environment, multiple resource providers share their resources and work together in order to accomplish a complex job. Shared resources include computation power, storage, memory, data, and special instruments such as telescope. Grid is a highly decentralized environment involving multiple autonomous administrative domains with its own policies such as scheduling, access cost and security. The aim of Grid is to utilize the idle CPU cycles of different contributing organizations for running complex scientific applications requiring days of computation. Efficient utilization of resources is primary goal of Grid environment. The service level agreement is negotiated between client and a provider for executing the job on high performance computing resources. The performance of the negotiation process highly depends on number of resources selected for the negotiation and process of the negotiation. Most of today grid applications consist of highly correlated tasks. The performance of the workflow application containing highly correlated tasks mainly depends on scheduling. This report proposes an improved SLA based Grid scheduling for workflow application using improved heterogeneous earliest finish time algorithm to select appropriate resource(s) to execute each task of workflow application and advance reservation based resource negotiation to get commitment of the resources towards the task. The proposed model schedule each task of workflow application not only keeps in mind the earliest finish time but also requirement of the user's job. The resultant effect of the proposed model decreases both execution times of the workflow application as well as rescheduling needed for each task. The implementation of the proposed model is done on Globus environment. The results are produced. The results indicate the improved performance in terms of number of successful SLA creations and execution time.