DDB分布式数据库系统 第一讲 Ddb-L1 introduction

1AbouttheInstructorName:GongZhiguoOffice:N415Phone:[phone]E-Mail:[e-mail]2FromFileProcessingtoDBMSFileofcurrentaccountsFileofsavingaccountsFileofcustomersProgram4customerinformationProgram1depositwithdrawProgram2transferProgram3printingstmtDBMSBANKDATABASEProgram4customerinformationProgram1depositwithdrawProgram2transferProgram3printingstmt3DDBS=Database+NetworkingThetechnologyofcomputernetworks,promotesamodeofworkthatgoesagainstallcentralizationeffortsandfacilitatesdistributedcomputingDistributeddatabasesystemtechnologyistheunionofwhatappeartobediametricallyopposedapproachestodataprocessing:DatabaseSystem,ComputerNetworktechnologiesAdatabasesystemaimsatintegratingtheoperationaldataofanenterprise,andtoprovideacentralizedandcontrolledaccesstothatdata4DistributedComputingSystemAdistributedcomputingsystemconsistsofanumberofautonomousprocessingelements(notnecessarilyhomogeneous)interconnectedbyacomputernetworkcooperateinperformingtheirassignedtasksWhatisdistributed?ProcessingLogicFunctionDataControlAllthesearenecessaryandimportantfordistributeddatabasetechnology5DistributedDBMSEnvironmentSite1Site2Site4Site3Site5Site6CommunicationNetwork6DistributedDatabaseSystemAdistributeddatabaseisacollectionofmultiple,logicallyinterrelateddatabasesdistributedoveracomputernetwork;storesdataonmultiplecomputers(nodes)overthenetworkandpermitsaccessfromanynodetothejointdataAdistributeddatabasemanagementsystem(DDBMS)isasoftwaresystemthatpermitsthemanagementofthedistributeddatabasesandmakesthedistributiontransparenttotheusers.7WhatisnotaDistributedDatabaseSystem?ADDBSisnota``collectionoffiles''thatcanbeindividuallystoredateachnodeofacomputernetworkfilesarenotlogicallyrelatednoaccessviacommoninterface8CentralizedDBMSonaNetworkdataresidesonlyatonenodethedatabasemanagementisnodifferentfromcentralizedDBMSremoteprocessing,singleservermultipleclientsSite1Site2Site4Site3Site5Site6CommunicationNetwork9MultiprocessorSystemsSharedMemory(SMP)SharedDiskSharedNothing(network)Sequent,SGI,SunVMScluster,SysplexTandem,Teradata,SP2CLIENTSMemoryProcessorsCLIENTSCLIENTS10DistributedDatabaseSystemTechnologyThekeyisintegration,notcentralizationDistributeddatabasetechnologyattemptstoachieveintegrationwithoutcentralizationDatabaseTechnologyComputersNetworksDistributedDatabaseSystemsIntegrationIntegrationWithoutCentralizationDistributedComputing11ExampleMultinationalmanufacturingcompany:headquartersinMacaumanufacturingplantsinNanningandKunmingwarehousesinZhongshanandDongguanR&DfacilitiesinBeijingDataandInformation:employeerecords(workinglocation)projects(R&D)engineeringdata(manufacturingplants,R&D)inventory(manufacturing,warehouse)12PromisesofDistributedDBMStransparentmanagementofdistributed,fragmented,andreplicateddataimprovedreliabilityandavailabilitythroughdistributedtransactionsimprovedperformancehighersystemextendibility13TransparencyTransparencyreferstoseparationofthehigher-levelsemanticsofasystemfromlower-levelimplementationdetails.FromdataindependenceincentralizedDBMStofragmentationtransparencyinDDBMS.IssuesWhoshouldprovidetransparency?Whatisthestateoftheartintheindustry?14ImprovedReliabilityDistributedDBMScanusereplicatedcomponentstoeliminatesinglepointfailure.Theuserscanstillaccesspartofthedistributeddatabasewith“propercare”eventhoughsomeofthedataisunreachable.Distributedtransactionsfacilitatemaintenanceofconsistentdatabasestateevenwhenfailuresoccur.15ImprovedPerformanceSinceeachsitehandlesonlyaportionofadatabase,thecontentionforCPUandI/Oresourcesisnotthatsevere.Datalocalizationreducescommunicationoverheads.Inherentparallelismofdistributedsystemsmaybeexploitedinter-queryparallelismintra-queryparallelismPerformancemodelsarenotsufficientlydeveloped.16EasierSystemExpansionAbilitytoaddnewsites,data,andusersovertimewithoutmajorrestructuring.Hugecentralizeddatabasesystems(mainframes)arehistory(almost!).PCrevolution(CompaqbuyingDigital,1998)willmakenaturaldistributedprocessingenvironments.Newapplications(suchas,supplychain)arenaturallydistributed-centralizedsystemswilljustnotwork.17DisadvantagesofDDBSsLackofExperienceNooperatingtruedistributeddatabasesystemsinexistenceComplexityDDBSproblemsareinherentlymorecomplexthancentralizedDBMSonesCostMorehardware,softwareandpeoplecostsDistributionofcontrolProblemsofsynchronizationandcoordinationtomaintaindataconsistencySecurityDatabasesecurity+networksecurityDifficulttoconvertNotoolstoconvertcentralizedDBMSstoDDBSs18ComplicatingFactorsDatamaybereplicatedinadistributedenvironment,consequentlytheDDBSisresponsibleforchoosingoneofthestoredcopiesoftherequesteddataforaccessincaseofretrievalsmakingsurethattheeffectofanupdateisreflectedoneachandeverycopyofthatdataitemIfthereissite/linkfailurewhileanupdateisbeingexecuted,theDDBSmustmakesurethattheeffectswillbereflectedonthedataresidingatthefailingorunreachablesitesassoonasthesystemrecoversfromthefailure19ComplicatingFactorsMaintainingconsistencyofdistributed/replicateddata.Sinceeachsitecannothave