计算机毕业设计外文翻译

毕业设计（论文）外文资料原文及译文专业班级学号姓名指导教师毕业设计（论文）外文资料原文1原文出处：DataCachingTradeoffsinClient-ServerDBMSArchitecturesbyMichaelJ.Carey/MichaelJ.Franklin/MironLivny/EugeneJ.Shekita,ComputerSciencesTechnicalReport#994,January1991DataCachingTradeoffsinClient-ServerDBMSArchitecturesInthispaper,weexaminetheperformancetradeoffsthatareraisedbycachingdataintheclientworkstationsofaclient-serverDBMS.Webeginbypresentingarangeoflock-basedofthewell-understoodproblemofreplicateddatamanagement.Wethenuseadetailedsimulationmodeltostudytheperformanceofthesealgorithmsoverawiderangeofworkloadsandsystemsresourceconfigurations.Theresultsillustratekeyperformancetradeoffsrelatedtoclient-servercacheconsistency,andshouldbeofusetodesignersofnext-generationDBMSprototypesandproducts.1.INTRODUCTIONWithnetworksofpowerfulworkstationsbecomingcommonplaceinscientific,engineering,andevenofficecomputingenvironments,client-serversoftwarearchitectureshavebecomeacommonapproachtoprovidingaccesstosharedservicesandresources.Mostcommercialrelationaldatabasemanagementsystemstodayarebasedonclient-serverarchitectures,withSQLqueriesandtheirresultsservingasthebasisforclient-serverinteractions.Inthepastfewyears,anumberofobject-orientedDBMSprototypesandproductshaveappeared,virtuallyallofwhicharebasedonclient-serverarchitectures.Comparedtorelationaldatabasesystems,thesesystemsgenerallytakeadifferentapproachwhenitcomestoclient-serverinteractions.Inordertooffloadprocessingtoclientworkstations,itiscommonforclient-serverinteractioninanOODBMStotakeplaceatthelevelofindividualobjectsorpagesofobjectsratherthanqueries.Prototypesbasedonobject-levelinteractionincludeOrionandO2,amongothers.Amongtheprototypesbasedonpage-levelinteractionsareObServerandthecurrentversionoftheEXODUSstoragemanager;theObjectStoresystemfromObjectDesignisanexampleofacommercialOODBMSproductbasedonpage-levelinteractions.Inarchitectureswheredatapagesorobjectsarethebasisforclient-serverinteractions,itispossibletocachedatainthelocalmemoriesofclientworkstationsforlaterreuse.Suchcachingcanreducetheneedforclient-serverinteraction,lesseningthenetworktrafficandmessageprocessingoverheadforboththeserveranditsclients.ItalsoenablesclientresourcestobeusedbytheDBMS,thusincreasingboththeaggregatememoryandtheaggregateCPUpoweravailablefordatabase-relatedprocessing.AnincreaseintheaggregatememoryoftheDBMScanreducetheI/Oloadontheserver,whileanincreaseintheaggregateCPUpoweravailabletotheDBMScanreducetheloadontheserverCPU(s).Dependingonthenatureoftheapplications--includingtheirbalanceofI/OandCPUdemands,theirlocalityofaccess,and毕业设计（论文）外文资料原文2theproportionalcostoftheirDBMSaccessesrelativetotheiroverallcomputationalrequirements--increasingtheaggregateresourcesoftheDBMScanresultinsignificantperformanceimprovements.Ofcourse,inapplicationswheretheresponsetimeisdominatedbythetimespentattheclientCPU,orwherealargefractionoftheDBMSisaccessedrelativetothesizeoftheclientbufferpool,theperformancebenefitsofcachingwouldbenegligible.Despiteitspotential,cachingisnotaperformancepanacea.Inordertoincorporatecaching,theDBMSmustincludeaprotocolthatensurescacheconsistency.Suchaprotocolmaybecomplextoimplement,itmayentailasignificantamountofprocessingoverhead,anditsimpactonsystemperformancemaybeworkload-dependent.Dependingonhowtheprotocolandtheworkloadinteract,thecacheconsistencyprotocolmightactuallyincreasetheloadontheserverand/ortheclientworkstationsduetoitsoverhead,particularlywhentherearealargenumberofclientworkstations.Anotherpotentialpitfall,whichdependsontheconcurrencycontrolschemeusedbytheprotocol,isthelatediscoveryofdataconflicts.Thus,thepotentialconsequencesofaddingcachingtoaclient-serverDBMSrangefromasignificantimprovementinperformancetoanotabledegradationinperformance.AnexampleofaworkloadwherecachingcanbehighlybeneficialistheSunEngineeringDatabaseBenchmark.Inthispaper,weexaminethedatacachingperformancetradeoffsdiscussedabove.Webeginbypresentingarangeoflock-basedcacheconsistencyalgorithmsthatresultfromrecognizingthatcacheconsistencyissimplyavariantofthereplicateddatamanagementproblemstudiedbydistributedDBMSresearchers.Forconcreteness,wefocusourattentiononsystemswhereclient-serverinteractionarepage-based.Thisapproach,alsoreferredtoastheblockserverapproach,wasshowntoperformwellforCAD-styledataaccesspatternsinarecentperformancesstudy.Also,ourworkwasmotivatedbyadesiretounderstandperformancetradeoffsinourownpage-based,client-serverstoragemanagers.Giventhissetofcacheconsistencyalgorithms,wethendescribeadetailedsimulationmodelthatwasdevelopedtostudytheirperformanceoverawiderangeofworkloadsandsystemresourceconfigurations.Theperformanceoftransaction-orientedcacheconsistencyalgorithmshasbeenexaminedinseveralrelatedcontexts.Theonlyotherclient-serverdatacachingstudythatweknowofisarecentsimulationstudyatHPLaboratories.Ourworkdiffersfromtheirworkinseveralways.First,weemployamuchmoredetailedmodelofbuffering,theimportanceofwhichwillbeclearfromourresults.Second,westudyabroaderrangeofDBMSworkloads.Ourworkisalsorelatedtostudiesofshared-diskarchitectures,includingth