Design of a Fault Tolerant Load Balancing Scheduler for Computational

Abstract

Grid is a collection of heterogeneous resources from a decentralized management. These resources are from large number of organizations and each resource with a local administration and scheduling policy are connected by wide area network. Due to these factors, strong variations prevail in the availability of grid resources and lead to increase in the probability of failures than in traditional parallel and distributed systems. An efficient fault tolerant scheduling approach is needed to schedule the jobs to the appropriate resources which can handle the failure nature of grid resources. Load balancing is required to improve the utilization of the resources and throughput. In this paper, a new scheduling approach is proposed that schedules the grid tasks by considering the failure rate of the resources in order to handle task failures. The proposed Fault tolerant Load Balancing (FTLB) algorithm increases the execution success rate and improves makespan. This algorithm also considers load of each resource while scheduling the jobs to increase resource utilization and throughput of the system. By considering user demand such as deadline for job completion, it improves user satisfaction. The analysis of the experimental results proves that the proposed FTLB algorithm has improved makespan, hit rate and average resource utilization than the existing algorithms.