A precedence based load balancing strategy for batch of DAGs for computational grid

Abstract

Load balancing on computational resources in a computational grid environment is an NP Hard problem. Therefore, a number of schemes can be proposed for the same under different constraints. Accordingly, a variety of techniques have been proposed in the literature but none of them can be treated as the best for all conditions and QoS parameters. A computational grid enables the users to execute their compute intensive jobs on the resources in which investing is either not wise or beyond their limit to invest. Grid resources can be used to meet the user's requirements in terms of QoS parameters if the scheduling strategy provides the efficient mapping between the software parallelisms available in the application on the hardware parallelism offered by the grid. This work presents a centralized precedence based load balancing strategy for a batch of jobs represented as Direct Acyclic Graph (DAG) having communication requirements. Performance evaluation is carried out by comparing the proposed strategy with other load balancing schemes viz. LRR and OLB. Experimental study reveals that load distribution on nodes is better than LRR and OLB. The strategy can be used for the jobs which are more interactive as it results in a lower response time for the jobs of the batch.