Research Of Job Scheduling Problem Of Grid Based On Performance Measured-Matrix

In the past ten years,large-scale scientific computing has been shifting from centralized mainframe to distribute and parallel pattern,recently this trend has extended to grid computing field.Grid is an important development trend for parallel and distributed computing technology nowadays,its goal is to achieve a large number of heterogeneous resources on a broad geographically sharing.Because of the character of distributed and shared,self-resembling,dynamic,diverse,autonomous and manifold management of grid system,which makes resource management technology and job scheduling technology to become the key research fields of current grid system.The key technology to improve the grid system efficiency is the job scheduling algorithm,how integrate various factors to making the scheduling algorithm more comprehensive is a challenging problem.At the grid system,good scheduling strategy can reduce the execution time of operatious,reduce costs,improve service quality and increase the throughput of grid system.This thesis describes the basic concepts, characteristics and application areas of grid,as well as grid resource management technology,discussed a number of traditional grid job scheduling algorithm,focused on the Min-min job scheduling.For the problem of load of Min-min algorithm is not balanced,this article based on the market economy model,achieves load balancing through the estimate grid resources' quoted price.Considered the grid system's complexity,in the performance Measured-Matrix's construction process,this article fully considered each kind of scheduling factor,which divided into three kinds of situations to carry on the quantification:grid resources handling ability,grid Quality of Service,the benefit function,finally presents these factors by the concrete quantification value for the grid job scheduling system.Through the mathematical method,obtains the performance Measured-Matrix.Then take the performance Measured-Matrix as the scheduling foundation,put forward the Measured-Matrix job scheduling algorithm. Finally,under the GridSim platform,carries on the simulation experiment to the Measured-Matrix job scheduling algorithm,simultaneously compares the results of the simulation experiment of Optimise Cost and Optimise Time job scheduling algorithm.By contrast,obtains that the Measured-Matrix job scheduling algorithm has big improvement in performance aspects of grid such as running time,load balancing,could better adapt to the dynamic and the extendibility characteristics.So the Measured-Matrix job scheduling algorithm has promoted the efficiency of job scheduling.