Research On Job Scheduling In Grid Resource Market

As a new technology in internet development,Grid computing gives a low cost solution for large-scale computation task,while making best use of the network resource.Although Grid has developed for years,there is still a big gap from general application. This thesis summarizes the current situation of Grid technology,and believes that Grid will be the next generation of interact.However,to get Grid into general application,a lot of problems need solving,and resource management is just one of the most important.This thesis summarizes the research of market-based Grid resource management,and believes that applying economic principles to Grid is deemed encouraging more people to participate in Grid and promising to improve the overall value provided by such participants. In this paper,a new market-based Grid resource management mode is given,which improves ease of use and applicability for large-scale Grid environment.Based on the different description of job,the resource scheduling model is divided into independent job scheduling and workflow scheduling.First,this thesis improves the HRED algorithm through resource selection mechanism and reliability quantizing,which make the scheduling of improved-HRED algorithm more effective and more appropriate for large-scale environment.Second,we also give a novel workflow scheduling model:GUM.It is a general, user-friendly,multi-criteria scheduling model,and performs better in universality, applicability for large-scale environment and user friendliness,which is also extensible in the number of criteria.By comparison analysis,we prove that GUM can meet user's needs better, and adapt to large-scale grid environment better.