Augmenting Hierarchical Load Balancing with Intelligence in Grid Environment

Scheduling independent tasks to homogeneous resources is an ineluctable issue to be dealt with. Load balancing of resources is a crucial matter of concern. This paper comes out with an enhancement of hierarchical load balancing algorithm. In this paper, to evaluate cluster imbalance, probability of deviation of average system load from average load of cluster is calculated and checked for confinement within a defined range of 0 to 1. The algorithm also compares the expected computing power of jobs with average computing power of clusters to allocate fittest resources to jobs. In addition to the load balancing and fittest resource allocation, the contribution of our algorithm is twofold. Our algorithm ensures that no cluster remain idle by employing random stealing and random pushing, whereby jobs are taken from other clusters and executed in cluster with empty queue such that the queue length remains within a fixed threshold. The contributions of augmented hierarchical load balancing with intelligence is that it reduces the makespan of algorithm execution together with balancing the overall system load and reduces the idle time of clusters.