Research On Network Bandwidth Constraint-based Algorithm For Grid Task Scheduling

Compared with the tradional parallel systems,the grid is a shared enterprising environment where is no central controLSo some newer challenges have arisen in the task scheduling.The main goal of grid task scheduling is to maximum its system throughput and to match the application's needs with the available computing resources.In addition,the matching of task with resources in a shared heterogeneous environment leads to concerns with Quality of Service(QoS).In the present,research on task scheduling algorithm for grid is at the beginning stage,it is difficult to propose a systematic scheduling solution to a heterogeneous,dynamic,scalable and large-scale grid.Moreover,centralized resources management has a serious effect on flexibility and fault-tolerance of task scheduling system.This paper proposes a network bandwidth constraint-based algorithm for grid task scheduling.This new algorighm is based on the general adaptive scheduling heuristics and an added network bandwidth guide component.It try to make a very significant modification that the jobs requiring high bandwidth are given higher priority.As a result,this measure prevents a situation,in which the jobs requiring high bandwidth have to wait for hosts occupied, which have been occupied by the jobs only requiring low bandwidth.This measure improves the throughput.It has been demonstrated in a simulated grid environment experiment which makes use of some technologies such as multithread and random number. In addition to grid task scheduling,some related topic,such as grid computing,task scheduling principle and architecture,performance prediction model,task deviding and parallelizing,basic scheduling service are involved.This study is a first attempt to support QoS in grid task scheduling.We have addressed only one-dimensional QoS issues.Many issues remain open.How to embed multi-dimensional QoS into task scheduling is still a topic of research.We have addressed only the concern of bandwidth of network.How to classify general QoS in a grid environment,choose scheduling frequency as well as adapt the algorithm to grid,still needs more deliberation.This research is financed by national natural science foundation(60402028), hubei province natural science foundation(2003ABA041) and Wuhan Chen Guang Project (20045006071-15).