建筑

数据仓库DataWarehouse:Fundamental&Technology第四章操作型数据存储OperationalDataStore第四章操作型数据存储4.1概述4.2什么是ODS？4.3OLTP-ODS-DW体系结构4.4ODS的创建4.5ODS实例4.6小结4.1概述数据仓库的主要贡献明确提出数据处理的两种不同类型:操作型处理，OLTP分析型处理，DSS将两种处理在实现中区分开来建立起OLTP-DW两层数据存储体系解决数据集成的问题TheArchitectedEnvironmentDataWarehouseMostgranularTimevariantIntegratedSubjectorientedSomesummaryDepartmentalParochialSomederivedSomeprimitiveTypicaldepartmentsIndividualTemporaryAdhocHeuristicNon-repetitivePC,workstationbasedOperationalDetailedDaytodayCurrentvaluedHighprobabilityofaccessApplicationoriented存在问题在很多情况下，OLTP-DW的两层体系结构并不能涵盖企业所有的数据处理要求。因为企业的数据处理虽然可以较为粗略地划分成操作型和分析型两部分，但有时，这两种处理之间并没非常明晰的界限。实际的数据处理往往是多层次的。也就是说，有些处理是操作型的，但不适合在操作型DB中进行，而又存在着一些分析型处理，但不适合在DW中进行。实例决策环境商场决策人销售部门的中下层管理人员决策问题某商品是否要进货？决策所需掌握的情况需要掌握以下信息储备是否充足？该商品近期销售情况如何？资金情况如何？其他商品的库存情况和销售情况如何？即要综合了解这些信息，才能做较为合理、可行的决策。在何处实现这些功能？方法1：放到分散的OLTP系统中去做不一定能得到每个部门的准确的一致信息，需要进行部门间的协调配合，工作量会很大。方法2：将其放在数据量巨大的DW中去处理显然会较费时，可能涉及许多不必要的数据检索。这类决策的特点特点不是在线事务处理也算不上是高层决策分析。属于日常管理和控制的决策问题企业中层的管理者经常要解决的、较大量的问题。前述两种解决方法都不太可行，该如何解决？另一种数据环境这种信息处理的特点引出了一种数据环境ODS，operationaldatastore操作型数据存储它是在OLTP-DW两层体系结构的基础上再增加一个层次ODS，从而形成OLTP-ODS-DW的三层数据存储体系。图示OperationalOLTPODSODSApplicationDataWarehouseDWApplication中间层次ODS中的数据内容两类数据一方面，它包含企业全局一致的、细节的、当前或接近当前的数据，可以进行全局在线操作型处理；另一方面，它又是一种面向主题、集成的数据环境，且数据量较小，适合于辅助企业完成日常决策的数据分析处理。第四章操作型数据存储4.1概述4.2什么是ODS？4.3OLTP-ODS-DW体系结构4.4ODS的创建4.5ODS实例4.6小结4.2什么是ODS4.2.1ODS的定义及其特点4.2.2ODS的功能与实现机制4.2.1ODS的定义及其特点定义ODS是用于支持企业日常的全局应用的数据集合保存在ODS中的数据具有四个基本特点面向主题集成的可变的数据是当前或接近当前的关于ODS因为数据面向主题，要求ODS中的数据在企业级上应该保持高度的一致性，所以必须对进入ODS数据进行转换和集成。区别于与分散在各个OLTP应用的数据：面向主题、集成化。区别与DW中的数据存放当前数据或接近当前的数据可以进行在线修改ReferenceMaterialsFourClassODSesFourclassesofODSFromthebeginning,itwaspredictedthattherewouldbedifferentclassesofODSsbasedonhowfastdatawasloadedintotheODSandthesourceoftheload.ItwaspredictedthattherewouldbefourclassesofODSs:ClassI.TransactionsweremovedtotheODSinanimmediatemannerfromapplications-inarangeofonetotwosecondsfromthemomentthetransactionwasexecutedintheoperationalenvironmentuntilthetransactionarrivedattheODS.Inthiscase,theendusercouldhardlytellthedifferencebetweenanactivitythathadoccurredintheoperationalenvironmentandthesameactivityasitwastransmittedintheODSenvironment.ClassII.ActivitiesthatoccurredintheoperationalenvironmentwerestoredandforwardedtotheODSeveryfourhoursorso.Inthiscase,therewasanoticeablelagbetweentheoriginalexecutionofthetransactionandthereflectionofthattransactionintheODSenvironment.However,thisclassofODSwasmucheasiertobuildandtooperatethanaClassIODS.ClassIII.ThetimelagbetweenexecutionintheoperationalenvironmentandreflectionintheODSisovernight.InaClassIIIODSthereisanoticeabletimelagbetweentheexecutionofthetransactionintheoperationalenvironmentandthereflectionofthetransactionintheODSenvironment.ThistypeofODSisrelativelyeasytobuild.ClassIV.AClassIVODSisonethatisfedfromthedatawarehousefromanalysiscreatedbytheDSSanalystinthedatawarehouseenvironmentandcondenseddowntoapointwheretheresultsoftheanalyticalprocessingfitcomfortablyintheODS.TheinputtotheODScanbeeitherregularorirregular.ThisclassofODSisveryeasytobuildaslongasthedatawarehousehasalreadybeenconstructed.AclassIVODSisonewheretheODSisfedfromanalysisdoneonthedatawarehouse.Thedecisionsupportanalystcreatestheanalysisandsynthesizesthedata.Oncesynthesized,thedataisfedintotheODSfromthedatawarehouse.AclassIVODSisquiteusefulforgettingtheresultsofthedatawarehouseprocessingclosetothecustomer.SinceonlineresponsetimeisnotanormalfunctionofthedatawarehousebutisanormalfunctionfortheODS,theODSservesastheperfectvehiclefortakingDSSdatawarehouseanalysisanddeliveringittothecustomer.ConclusionAsarule,ClassIODSsarerare.ThenormalcaseforanODSisatypeII,IIIorIV.ClassIODSsaredifficulttoconstructandevenmoredifficulttooperate.ThereneedstobeaverystrongbusinesscaseforaClassIODS.ReferenceMaterialsWhat’sODS?What’sODS?Anoperationaldatastore(ODS)isatypeofdatabaseoftenusedasaninterimareaforadatawarehouse.Unlikeadatawarehouse,whichcontainsstaticdata,thecontentsoftheODSareupdatedthroughthecourseofbusinessoperations.Forwhat?AnODSisdesignedtoquicklyperformrelativelysimplequeriesonsmallamountsofdata(suchasfindingthestatusofacustomerorder),ratherthanthecomplexqueriesonlargeamountsofdatatypicalofthedatawarehouse.DifferencefromDWAnODSissimilartoyourshorttermmemoryinthatitstoresonlyveryrecentinformation;incomparison,thedatawarehouseismorelikelongtermmemoryinthatitstoresrelativelypermanentinformation.ClassIIIIntheearly1990s,theoriginalODSsystemsweredevelopedasareportingtoolforadministrativepurposes.Theywereusuallyupdateddailyandprovidedreportsaboutbusinesstransactionsforthatday,suchassalestotalsorordersfilled.ThistypeofsystemisnowreferredtoasaClassIIIODS.ClassIIWithchangesintechnologyandbusinessneeds,theClassIIODSevolvedtotrackmorecomplexinformationsuchasproductandlocationcodes,andtoupdatethedatabasemorefrequently(perhapshourly)toreflectchanges.ClassIClassIODSsystemsarosefromthedevelopmentofcustomerrelatio