Performance based Resource Scheduling in Diverse Multi Cluster Grid Environment

Abstract

Resource scheduling is an important aspect in grid computing where the users (consumers) and resources (clusters) are distributed geographically across different administrative domains. This research addresses the establishment of a resource scheduling algorithm in diverse multi-cluster grid architectures, like centralized, decentralized and hybrid. In centralized architecture, a single scheduler is responsible for monitoring and maintaining the details of all the clusters present in the grid environment. Because of the single scheduler, the total time the job spends in the grid system is very high when the load increases. In order to improve the performance of the centralized grid system further, fault tolerant features is be incorporated in grid resource scheduling. The resource fault index based checkpoint set strategy is integrated into the centralized resource scheduling algorithm. A decentralized resource scheduling and load balancing algorithm is proposed. The decentralized dynamic algorithm shows a marked improvement in the reduction of the total time the job spends in the grid system, and also increases the number of jobs completed within the specified deadline, better than the centralized grid architecture. This is because the selection of the resource for job execution is done locally for every cluster. The performance of the proposed centralized, decentralized and hybrid resource scheduling algorithms are compared. The percentage of improvement of one algorithm over other algorithms is discussed. The overhead involved in each architecture model is also analyzed. The performance of the algorithm depends on the parameters, the machine configurations and the workload of the resource. The proposed resource scheduling algorithms are shown to be robust in terms of the selection of suitable resources, reduction in the total time, and the communication overhead, an improved job completion rate for the user community and improved resource utilization. newline newline newline