Decentralized Computation and Communication Intensive Task Scheduling Algorithm for P2P Grid

Abstract

Most of earlier grid scheduling algorithms were based on centralized scheduler. Relying on central scheduler yields not only central point of failure, also, it is not possible because of scalability and political issues in present day gigantic grid systems. Hence, meta-schedulers came into limelight. However many authors recognizes limitations of hierarchical grid scheduling and proposed peer-to-peer (P2P) techniques, which have potential for grid scheduling. In this paper, new decentralized scheduling algorithm is proposed for P2P grid systems. In this method, an independent task sovereignly selects a most suitable grid node based on local information of immediate neighbors. A vital feature of this method is that it can schedule both computation intensive and communication intensive tasks to make grid system's workload balanced.