校园计算网格环境下作业管理的研究与实现

《校园计算网格环境下作业管理的研究与实现》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。

1、西北工业大学硕士学位论文校园计算网格环境下作业管理的研究与实现姓名：侯正雄申请学位级别：硕士专业：计算机应用技术指导教师：周兴社20050101西北_1业大学硕卜学位论文ABSTRACTThemainpurposeofgridistosharetheexistingcomputingresourcestoworkcooperatively.TosolveproblemsbasedongridisgridcomputingGlobusisatypicaltechnologyofgridcomputingandthedefactstandard.Itdoesthekeytec

2、hnologyresearchofgridcomputingtohelplayoutandbuildbiggridcomputingtestbed.Ittriestoimplementthegridinfrastructuretoprovidetheusersuniversal,reliableandconsistentaccesstothehighperformancecomputingresourcesremotelyandtransparently.Thegirdusersusuallyusethegridresourcesbygridjobsandapplicat

3、ions.Therefore,howtoscheduleandmanagethegridjobsefficientlytomeettheusersapplicationrequirementsisoneofthekeytechnologiesingirdcomputingresearch.Thisthesisintroducesthegridandit'sarchitecture.Itanalyzesthedevelopmentofthejobmanagementsystem.So,wehaveadefiniteconceptofthejobmanagementingri

4、dcomputing.Weconcretelydotheresearchofgridjobscheduling,includingschedulingpolicyandalgorithm,thequeuingscheduling，advancedreservationandparallelcomputingprogrammodel.Themodeloftherelationbetweenglobalschedulingandinternalschedulingisalsodesigned.Then,wediscussthejobmonitoring，controlandm

5、anagementfunctions,suchasjobmonitoringandcontrol,coarsegrainjobdisassembleandacquiringtheresults.Theuserinterfaceofthejobmanagementsystemisalsodiscussed.Weimplementtheprototypeoftheserviceorientedcampuscomputationalgridandit'sjobmanagementsystem.Basedonagents,thelayergridresourcetreemodel

6、isadopted.ProposeaglobalMCTschedulingalgorithmbasedonthecreditofthegridnodes.Realizetheinterfacebetweenthegridmiddlewareandlocalresourcemanager.Monitorandcontrolthewholegridsystemwiththehelpoflocalmanagementsystem.Atlast,weimplementtheunifiedresourceandusersinformationmanagement,unifiedjo

7、bmanagementandsafeauthentication.Withtheconsiderationofthepracticalsituationonthecampus,wedeploythehighperformancecomputingserviceorientedcampuscomputationalgird.KeyWords:campuscomputationalgrid;jobmanagement;jobscheduling;jobmonitorandcontrol;jobservice;highperform