存储过程英文原文与中文译文.docx
- 文档编号:7299334
- 上传时间:2023-01-22
- 格式:DOCX
- 页数:26
- 大小:226.03KB
存储过程英文原文与中文译文.docx
《存储过程英文原文与中文译文.docx》由会员分享,可在线阅读,更多相关《存储过程英文原文与中文译文.docx(26页珍藏版)》请在冰豆网上搜索。
存储过程英文原文与中文译文
DataProvenanceSupportinRelationalDatabasesforStoredProcedures
WinlyJurnawanandUweRöhm
SchoolofInformationTechnologies,UniversityofSydney,
SydneyNSW2006,Australia
Abstract
Theincreasingamountsofdataproducedbyautomatedscientificinstrumentsrequirescalabledatamanagementplatformsforstoring,transformingandanalyzingscientificdata.Atthesametime,itisparamountforscientificapplicationstokeeptrackoftheprovenanceinformationforqualitycontrolpurposesandtobeabletore-traceworkflowsteps.Relationaldatabasesystemsaredesignedtoefficientlymanageandanalyzelargedatavolumes,andmodernextensibledatabasesystemscanalsohostcomplexdatatransformationsasstoredprocedures.However,therelationalmodeldoesnotnaturallysupportdataprovenanceorlineagetracking.Inthispaper,wefocusonprovidingdataprovenancemanagementinrelationaldatabasesforstoredprocedures.Ourapproach,calledPSP,leveragestheXMLcapabilitiesofSQL:
2003tokeeptrackofthelineageofthedatathathasbeenprocessedbyanystoredprocedureinarelationaldatabaseaspartofascientificworkflow.Weshowhowthisapproachcanbeimplementedinastate-of-the-artDBMSanddiscusshowthecapturedprovenancedatacanbeefficientlyqueriedandanalyzed.
Keywords:
Provenance,StoredProcedure,RelationalDatabase.
1Introduction
Scientificresearchiscurrentlyexperiencingadigitalrevolution.Computer-supportedscientificinstrumentssuchasnext-generationDNAsequencersorradiotelescopescanconductlargeexperimentalseriesautomaticallyina24x7settingandgeneratemassiveamountsofdata.Forinstanceinbioinformatics,the1000GenomesProjectemploysnext-genDNAsequencingtechnologythatgeneratesapproximately75Terabyteofdataweekly–injustoneofthreeparticipatinglabs.Anotherexample,istheexponentialgrowthoftheNCBIGenBankdatasince1998whichgrewtomore
than56billionbasepairs(approx56Gb)by2005.AndinPhysics,aprominentexampleistheLargeHadronCollider(LHC)projectthatexpectsdatavolumestohittheExabyte(1018B)scale.
Theselarge-scalescientificexperimentsdependonareliableandscalableproductionenvironment.Forexample,the1000Genomeprojectoperatesmorethan30next-genDNAsequencersinparallel,producingTBofrawsequencingdataautomaticallywhichhastobefurtherprocessed,analyzedandarchived.Giventhehighexperimentaldatathroughputgeneratedbyvarioussequencermachines,supervisedbymanyscientistsandrunningin24x7operationmode,itisimpossibletokeeptrackontheexperiment
processmanually.Thetraditionalroleoflablogbookrecordingdoesfailinsuchanenvironmentbecauseitisoutperformedbythethroughputofthedata.Hence,acomputersupporttokeeptrackoftheseprocessesisneeded;thissupportisoftenreferredasdataprovenance.
Provenanceinbriefisthehistoryofadatawhichiscollectedthroughoutits
lifetime.Thehistoryreferstotheprocessthatproduced,manipulatedandaffectedthedata.E-scienceresearchusesprovenanceinformationinvariouswaysdependsonthepurposeofresearchandtherequirementsofhistoricalinformation.Forinstanceinhumangenomesequencing,theamountofchemicalusedinweblabiscrucialtoensurethereliabilityofthesequenceresults.
Therehavebeenseveralattemptstomanagescientificdatawithrelational
databasesandtoimplementscientificprocessesinsideaDBMSusingexisting
extensibilityfeatures.Forinstance,dbBLASTforalignmentsearchesonlargegenedatabases,SDSSinastronomyresearch,andSciDBforthePhysicscommunity.RecentlylookedintosupportingawholeDNAsequencingpipelineforhigh-throughputgenomicsinsidearelationaldatabaseusingCLR-basedstoredprocedures.Theseattemptsindicatethatlarge-scaleexperimentalresearch,whichisstillmostlyfile-centricorganized,willslowlymovetowardsdatabase-centricapproaches.Thisphenomenonbringsnewchallengetothedatabasecommunity,becausethereisnocommonmodelforcapturingandrepresentingdataprovenancefordatabase-centrice-scienceresearches.
Inthispaper,proposeaconceptualprovenancemodeltotracktheworkflowprovenancefore-sciencetoolswhichisimplementedinrelationaldatabase.WeimplementedProvenanceforStoredProcedure(PSP)whichistheproof-of-conceptforourprovenancemodel.ItisimplementedinsidetherelationaldatabaseusingCLRstoredprocedurewithprovenancedatarepresentedinSQL/XMLformat.Inthisinitialwork,wefocusontheprovenancefordatawhichismanipulatedbystoredprocedure.
Ourmaincontributionsare:
1.Aconceptualprovenancemodelforrelationaldatabasesthatconsistsofthreemaincomponents:
data,agentandprocess.
2.WeproposeanXML-basedrepresentationforprovenancedatainrelationaldatabases.XMLformatischosentoaccommodatetheversatilityoftheprovenancedatarequirementwhichchangeoverthetime.
3.WeimplementedtwovariantsofPSPasproof-of-concept:
NaïvePSPandCentralizePSPwhichdiffersintheprovenancedatastoragescheme.WeevaluatedbothapproachesusingdbBLASTasagentandthehumangenomeasdataset.
Thestructureoftheremainingpaperisasfollows.Section2presentstheotherworksonprovenancewhicharerelatedtoourresearch.TheconceptualfoundationofourprovenancemodelinrelationaldatabaseispresentedintheSection3.InSection4,wepresenttheimplementationofPSPandinstanceofourapproaches.Section5presentstheexperimentalresultanditsanalysis.Andinsection6and7presentourfutureworkandconclusions.
2RelatedWork
Therearemanyexistingprovenanceapproacheswhichareimplementedontopofvarioustechnologiessuchasscriptingarchitectures,service-orientedarchitecture,andrelationaldatabasearchitecture.Inthissection,wefocusmoreontheprovenanceapproachesthataremostlyrelatedtorelationaldatabasearchitectures.
Therearetwomajorcategoriesofprovenancewhicharekeptinscientificresearch,thedataflowprovenanceanddataworkflowprovenance.Thedataflowprovenancefocusonrecordinghoworwherethedatahascopiedandmovedthroughoutdatabases.Similarly,describethedataflowprovenanceas“whereprovenance”whichmainlykeeptrackonthesourcelocation(s)ofcurrentdata.Dataflowprovenancecapturesandstoresinformationsuchas,thesourceofdata,typeofoperationdonetoretrievethedata(i.e.copy,move,insert),whoaccomplishedtheoperations,datesandversionofthedata.Ontheotherhand,thedataworkflowprovenancefocusesmoreonrecordingthehistoriesofprocessesthathasbeendonetoparticulardatauntilcurrentstage.Itcouldbeconsiderastherecipetoproducethecurrentstateofthatparticular
data.Workflowprovenanceisusuallyusedinresearchexperimentsandexplorations,wheretheprovenanceinformationgivesthefirmverificationsandexplanationsofresultsproduced.However,scientistusesprovenanceinformationinvariouswaysdependsonthepurposeofresearchandtherequirementsofhistoricalinformation.Moreoverthisrequirementofprovenanceinformationwillvaryoverthetime.Followingaresomeoftherelatedworkoftheexistingworksthatarerelatedtoourresearch.
CPDB(CopyPasteDatabase)istheimplementationofprovenanceconceptonwhereprovenancewhichwasintroducedin.ThemainfocusontheimplementationofCPDBisontheprovenanceforcurateddatabasewheremostofitscontentarecopiedorderivedfromothersources.Thus,CPDBfocusondataprovenancewhichrecordsthesourceofthedatainsteadoftheworkflowprovenancethatrecordtheseriesofprocessoreventthataffectthedata.TheactualprovenanceinformationisrepresentedintheXMLdataandstoredseparatelyfromtheactualdata,inCPDBitisstoreseparatelyin
auxiliarytable.CPDBisstillimplementedoutsidetherelationaldatabaseusingJavaapplication.
AnotherimplementationisREDUXwhichisaprovenancemanagement
systemwhichcapturestheprovenanceinformationfortheworkflowwhichisbuiltbased-ontheWindowsWorkflowFoundation(WinWF).REDUXstoresitprovenanceinformationintheRDBMSwhichsupportbetterdatamanagementanddataqueries.Howeveritisstillanimplementationoutsidetherelationaldatabasesystemwhichmakesuseofrelationaldatabasetostoretheprovenancedata.AnotherfeatureprovidebyREDUX,issmartreplaywhichenablestheusertoreplayonwhathashappenedtoaparticulardatabefore,thehistoryofthedata.
ThereareimplementationsofprovenancecapturinginthedatabasewhichareDBNotesandTrio.DBNotesisoriginallytheimplementationofannotationmanagementforadatainrelationaldatabases.Theannotationsofannotateddataarepropagatedwhentheparticulardataistransformed.DBNotestake
theadvantageoftheannotationsthatarepropagatedduringtransformationastheprovenancetrace.Theauthorsalsodefineprovenanceannotationswhereallthedataisannotatedwithitsaddress.Hence,astheannotationsarepropagated,oneoftheannotationsinformationcontainstheoriginaladdressofthedata.DBNotescanbecategorizedaswhereprovenance,ittracestheoriginofdatainsteadoftheworkflow.Ontheotherhand,Trioisanimplementationofprovenancetrackingwhichtracetheinformationforviewdataanditstransformationinthedatawarehouseenvironmentthroughqueryinversionmethod.TriooperatesintheRDBMSenvironmentwherethedataarequeried,copied,moved,etc.Itisadataprovenancebecauseitconcernsmoreonkeepingtrackofthesourceororiginofparticulartupleinaviewtable.Itusestheinversionmodeltoautomaticallytracethesourceofthedataforsetoftuples
- 配套讲稿:
如PPT文件的首页显示word图标,表示该PPT已包含配套word讲稿。双击word图标可打开word文档。
- 特殊限制:
部分文档作品中含有的国旗、国徽等图片,仅作为作品整体效果示例展示,禁止商用。设计者仅对作品中独创性部分享有著作权。
- 关 键 词:
- 存储 过程 英文 原文 中文 译文
![提示](https://static.bdocx.com/images/bang_tan.gif)