外文原文DISTRIBUTEDDATABASEDEVELOPMENTCENTRALIZEDDATABASESHAVEBEENINLARGESCALEUSEINTHEDATAPROCESSINGINDUSTRYSINCETHEEARLY1970SDISTRIBUTEDDATABASESAREVERYMUCHMORESOPHISTICATEDTHANCENTRALIZEDDATABASES,ANDTHEYINVOLVECOMPLEXSOFTWARETOHANDLEPROBLEMSSUCHASCONCURRENCYCONTROL,DEADLOCKPREVENTIONANDTHEMAINTENANCEOFDATAINTEGRITYDURINGMULTISITEUPDATESITISONLYRECENTLYTHATTHESESOFTWAREPROBLEMSHAVEBEENUNDERSTOOD,ANDPRODUCTSWHICHINCORPORATETHISSOFTWAREHAVEONLYVERYRECENTLYSTARTEDTOAPPEARONTHEMARKETTHEPRESENTGENERATIONOFDDBSMAINLYINVOLVEREPLICATIONOFDATABASESATSEVERALSITES,RATHERTHANDISTRIBUTIONOFINDIVIDUALDATABASESTHENEXTGENERATIONOFDDBS,THOUGH,ISLIKELYTOINVOLVEDISTRIBUTIONOFDATAOVERSEVERALGEOGRAPHICNODES,INORDERTOPLACEDATAASCLOSEASPOSSIBLETOTHELOCATIONATWHICHTHEHIGHESTRATEOFACTIVITYTAKESPLACEWITHTHESEDATABASESITISLIKELYTHATMULTISITEUPDATESWILLBEMORECOMMONLYUSEDTHANATPRESENTTHISWILLINVOLVEVERYCAREFULPREPARATIONOFRECOVERYPROCEDURESINORDERTOENSURETHATTHERISKOFLOSSOFDATAINTEGRITYISDRIVERDOWNTOANABSOLUTEMINIMUMINSOMEINSTALLATIONSITWILLALSOBENECESSARYTOCAREFULLYMONITORBOTHTHENETWORKANDTHEDDBSONANONGOINGBASIS,INORDERTOENABLETHEOPERATIONSSTAFFTOREACTRAPIDLYTOANYSITUATIONINWHICHDATACORRUPTIONHASTAKENPLACESOMEWORKHASALREADYBEENDONEINTHISAREA,USINGEXPERTSYSTEMSTOHELPINDETERMININGWHETHERAGIVENPROBLEMISACOMMUNICATIONSPROBLEMORADATABASEPROBLEMHOWEVER,THEREISANEEDFORTHESEDIAGNOSTICTECHNIQUESTOMATURETOAPOINTWHERETHEYCANBEUSEDONANEVERYDAYBASISBYCOMPUTEROPERATIONSSTAFFINTHEIMMEDIATEFUTURE,THEMOSTINTENSIVEDEVELOPMENTWORKWILLBEINTHEAREAOFTHEOPTIMIZERTHISISTHEMODULEUSEDBYTHEDDBINORDERTOMINIMIZECOMPUTINGANDCOMMUNICATIONSCOSTSAREGENERALLYHIGHCOMPAREDWITHPROCESSINGCOSTSWITHTHEEMPHASISBEINGVERYMUCHONTHELATTERASCOMMUNICATIONSCOSTSAREGENERALLYHIGHCOMPAREDWITHPROCESSINGCOSTSASOPHISTICATEDOPTIMIZERWILLTAKEACCOUNTOFCOMMUNICATIONSCOSTS,DISCI/OANDPROCESSINGCOSTS,ANDWILLALSOEVALUATESEVERALWAYSOFTRANSPORTINGDATAACROSSTHENETWORKTOSATISFYTHEREQUIREMENTSOFAMULTISITETRANSACTIONITWILLALSOEVALUATESEVERALACCESSPATHSINTOTHEDATBASE,ANDINSOMECASESWILLSTOREINFORMATIONABOUTITSCOMPARISONSOF,DIFFERENTACCESSSTRATEGIESFORFUTUREUSEASAGENERALRULE,OPTIMIZATIONPROBLEMSBECOMECOMPLEXWHENTHENUMBEROFNODESINTHEDDBEXCEEDSFIVEANOTHERIMPORTANTFUTUREDEVELOPMENTISLIKELYTOOCCURINTHEAREAOFSYSTEMARCHITECTURESASCHOOLOFTHOUGHTHASRECENTLYDEVELOPEDWHICHADVOCATESTHATTHEMOSTATTRACTIVEMODELONWHICHTOBASEADDBISTHECLIENTSERVERMODELINTHISMODEL,THE“CLIENTS”THEDATAUSERS,ARESEPARATEDFROMTHE“SERVERS”THEDATAPROVIDERSANDARELINKEDTOTHEMBYAHIGHSPEEDLANTHEMODELHASTHEATTRACTIONTHATCLIENTSANDSERVERSCANEASILYBEADDEDINAMODULARFASHION,ANDTOSOMEEXTENTDIFFERENTHARDWARETYPESCANBEMIXEDHOWEVER,ITISVERYDEPENDENTONRELATIVELYSOPHISTICATEDFUTUREDEVELOPMENTSINLANTECHNOLOGY,PARTICULARLYIFHETEROGENEOUSCOMBINATIONSOFHARDWAREARETOBESUPPORTEDTHEMOSTNOTABLEPRODUCTSTHATCURRENTLYREFLECTTHECLIENTSERVERMODELARESYBASE,INGRES,ANDTANDEMSNONSTOPSQLDDBSARECURRENTLYTHESUBJECTOFCONSIDERABLEINTERESTAMONGSEVERALHARDWAREANDDBMSSUPPLIERS，ANDASSUMINGTHATTHEPROBLEMSINVOLVEDINRECOVERY,DIAGNOSTICS,OPTIMIZATINGANDLANARCHITECTURECANBESOLVED,THEDISTRIBUTEDDATADASEAREALOOKSSETTOBEAMAJORAREAOFGROWTHFORTHECOMPUTERINDUSTRYINTHE1990S。DATADASESDIVIDEINTOCENTRALSERVERSANDLOCALFRONTENDSWEVEHEARDALOTMAYBETOOMUCHABOUTTHECHANGESTAKINGPLACEINTHEDATABASEMARKETYOUKNOWTHELITANYSQL,DATABASESERVERS,CLIENT/SERVERRELATIONSHIPS,ETCLETSSTARTBYREVIEWINGTHEBASICEVOLUTIONOFPERSONALCOMPUTERDATABASESDATABASESONPERSONALCOMPUTERSHAVETYPICALLYBEENDIVIDEDINTOTWOMINGROUPSFLATFILEMANAGERSWORKWITHASINGLEFILEATATIMEANDAREUSUALLYMENUDASED,EASYTOUSE,ANDAIMEDATANENDUSERWHOKEEPSTRACKOFHISORHEROWNDATA,SUCHASATELEPHONEORADDRESSLISTRELATIONALDATABASECANWORKWITHMULTIPLEFILESATATIMEANDHAVETYPICALLYBEENAIMDATCONSULTANTSORINHOUSESUPPORTSTAFFWHOBUILDAPPLICATIONSINTHEPASTFEWYEARS,THESEDISTINCTIONSHAVEBEGUNTOBREAKDOWNONTHEONEHAND,THEFLATFILEMANAGERSHAVEGOTTENALOTMOREPOWERFULPROGRAMSANDHAVEBEGUNTOAPPROACHRELATIONALDATABASESINTERMSOFBASICCAPABILITIESRELATIONALDATABASESHAVEBEENCHANGINGALSOPROGRAMSLIKEPARADOXAREDESIGNEDSOTHATITSSIMPLEFORAUSERTOPERFORMADHOCQUERIESANDPRODUCTSLIKEDATAEASELETYOUDESIGNMULTIFILEAPPLICATIONSWITHLITTLEORNOPROGRAMMINGATTHESAMETIME,HOWEVER,OTHERDATABASEPRODUCTSAREEMPHASIZINGPOWERANDMULTIUSERFEATURESOVEREASEOFUSETHISISPARTICULARLYTRUEODTHEREALLYPOWERFULDATBASEENGINESORIGINALLYDESIGNEDFORMINICOMPUTERSANDMAINFRAMESTHATARENOWOFFEREDANDPROMOTEDINPCVERSIONSTHEMOSTSUCCESSFULEXAMPLEOFTISHASBEENORACLE,ANINCREDIBLYPOWERFULDATABASEENGINE,ANDONEDESIGNEDFORMULTIUSERENVIRONMENTS,BUTLACKINGINTHEEASEOFUSEFEATURESTHATMOSTPCUSERSEXPECTMARKERLEADERDBASEHASBEENTRYINGTOCHANGETOMATCHALLTHESEEXPECTATIONSDBASEIVADDSAFRONTENDTHATMAKESITEASIERTOUSE,AQUERYBYEXAMPLEINTERFACEFORADHOCQUERIES,ANDMOREPOWERINTHEFORMOFMOREADVANCEDINDEXINGANDRELATIONALITHYALLOFTHISISCOMINGTOAHEADASMOSTDATABASEVENDORSRECOGNIZETHATWEREMOVINGTOAWORLDINWHICHMANYBUSINESSUSERSOFDATABASESYSTEMSARECONNECTEDTOALOCALAREANETWORKANDPERHAPSVIAGATEWAYSTOMINICOMPUTERSANDMAINFRAMESCONSIDERHOWDATABASESHAVEPROGRESSEDWHENWEHADJUSTSINGLEUSERS,DATABASESWERECONTAINEDONLOCALHARDDISKSASWEMOVEDTOWARDMULTIUSERDATABASES,THENEXTSTEPWASTOHAVETHEDATABASEITSELFRESIDEONACENTRALFILESERVER,ASISNOWPOSSIBLEWITHALLOFTHEMAJORPCDATABASESTHOUGHTHENETWORKSINVOLVEDCANBESERVERBASEDSUCHASNOVELLNETWAREORPEERTOPEERSUCHAS10NET,THEINFORMATIONISMERELYSTOREDONNETWORKDISKSTHELOCALWORKSTATIONORPERSONALCOMPUTERDOESALLTHEPROCESSINGTHATSTHECASETODAYWITHPRODUCTSSUCHASDBASEPLUSORWITHTHELANPACK,RBASE,ANDPARADOXALLTHISISNTTOOSOPHISTICATED,BUTITHASWORKEDBUTNOW,ASWEMOVETOWARDMOREPOWERFULPCSANDMORESOPHISTICATEDNETWORKOPERATINGSYSTEMS,THEVENDORSHAVEBEGUNTOTAKEANEWAPPROACHTHEBIGDIFFERENCEISTATNOWVENDORSSEEDATABASESASTWOPARTSTHEFRONTEND,WHICHCONTROLSHOWTHEDATAAREDISPLAYEDONSEREENANDNOWTHEUSERINTERACTSWITHTHEDATA,ANDTHEDATABASEENGINE,WHICHACTUALLYMAINTAINSTHEDATAANDPERFORMSOPERATIONSSUCHASQUERIES,SORTS,ANDINDEXESTHISISOFTENREFERREDTOASACLIENT/SERVERAPPROACH,WITHTHEINDIVIDUALPERSONALCOMPUTERSACTINGASCLIENTSTOACENTRALIZEDDATABASESERVERCLIENT/SERVERMEANSTHISPRODUCTSOPERATIONISINTRINSICALLYLINKEDTOTHENETWORKSQLSERVERISASOFTWAREDATABASE“ENGINE”THATRUNSEITHERONANETWORKSERVER,ONADEDICATEDNETWORKWORKSTATION,ORANONDEDICATEDWORDSTATIONRUNNINGOS/2THISENGINEPERFORMSSOCALED“BACKENDPROCESSING“FUNCTIONS,WHICHINCLUDEMANAGINGTRANSACTIONS,SECURITY,SORTING,ANDREADINGANDWRITINGDATATHEIDEABEHINDCLICENT/SERVERARCHITECTUREISTOSPREADTHEPROCESSINGLOADWHEREITBELONGSTHESERVER,FOREXAMPLE,HANDLESSPROCESSINGFORTASKSTHATAPPLYTOEVERYONESECURITY,ACCESSINGCOMMONDATA,ETCTHEFRONTENDAPPLICATIONSISLOCALIEDWHERETHENEEDEXISTSNEEDLESSNETWORKTRAFFICISTHUSCUTOUTTHISISSUPPOSEDTORESULTINLESSNETWORKPERFORMANCEDEGRADATIONASMOREPEOPLEUSETHESQLSERVERONEMARKETRESEARCHCOMPANYSAID“ASPRODUCTSCONFORMINGTOTHENEWCLIENT/SERVERARCHITECTUREAREROLLEDOUT,VENDORSWILJOCKEYFORMARKETPOSITION,WHILEUSERSARELIKELYTOSITOUTTHEINITIALPRODUCTPUSH”IBMINTRODUCEDOS/2EXTENDEDEDITION,WHIHADDSTOOS/2ITSOWNDATAMANAGER,ESSENTIALLYAMICROCOMPUTERVERSIONOFITSPOPULARDB2DATABASEFORMAINFRAMES,ANDASHTONTATEANDMICROSOFTBEGANSHIPPINGSQLSERVER,THEIRDATABASESERVER,WHICHISBASEDONANENGINEFROMSYBASEBOTHPRODUCTSCANBEADDRESSEDINSEVERALWAYSBYWRITINGCODEINALANGUAGESUCHASC,BYSENDINGTHEDATABASESQUERIESUSINGSQL,ORBYUSINGANOTHERDATBASEPROGRAMTATSERVESASAFRONTENDTHESEPACKAGESARENTDESIGNEDTOBEFRONTENDSOREVENDATABASEPROGRAMMINGLANGUAGESINTHEMSELVESORACLEHASBEGUNPOSITIONINGITSDATABASEINTHESAMEWAY,ASANENGINETHATCANBEADDRESSEDBYAVARIETYOFFRONTENDSALREADY,ORACLEHASINTERFACESAVAILABLEFROM123,DBXL,ANDHYPERCARDONTHEMACINTOSHSIMILARAPPROACHESARETAKENBYOTHERDATABASEVENDORSWHOCOMEFROMTHEMINICOMPUTERWORLDMICRORIMHASPROPOSEDAVARIATIONWITHITSNEWPRODUCTLINE,CODENAMEDATLAS,WHICHISDEIGNEDTOALSOMANAGEANDCOORDINATEDATAMAINTAINEDINMANYOTHERFORMATS,BYOTHERDATABASESERVERS,ANDONMULTIPLEMACHINES,INCLUDINGDATAFROMRBASE,DBASE,ANDEXTENDEDEDITIONASWELLASDATAINMOSTMINICOMPUTERDATABASES,SUCHASORACLE,INGRES,INFORMIX,ANDUNIFYTHEATLASENGINEISALSODESIGNEDTOBEACCESSEDVIAMULTIPLEFRONTENDSWHICHMICRORIMCALLS“SURFACES”INSUCHANENVIRONMENT,THEDATABASEENGINEWHICHCOULDBEOS/2EXTENDEDEDITION,SQLSERVER,ORACLE,GUPTASSQLBASE,ORSOMETHINGELSEENTIRELYRESIDESONACENTRALDATABASESERVER,USUALLYACOMPUTERANDHARDDISKDEDICATEDTOTHATPURPOSETHEINDIVIDUALENDUSERSCANTHENUSEAVARIETYOFDIFFERENTFRONTENDSORAPPLICATIONSDEVELOPEDINAVARIETYOFOTHERDATABASELANGUAGESTHEREWILLBEDIFFERENCESINHOWTHEVARIOUSFRONTENDSACTUALLYCOMMUNICATEWHICHTHEVARIOUSENGINESMOSTOFTHEPRODUCTSUSESOMEVARIATIONONSQL,ALANGUAGEFORQUERYINGDATABASES,POPULARIZEDBYIBMSDB2SOMEPRODUCTSWILLSENDACTUALSQLQUERIES,SOMEWILLCOMMUNICATETHROUGHAPISAPPLICATIONPROGRAMINTERFACES,ANDSOMEWILLSENDMESSAGESINTHESERVERSNATIVECODE,BUTTHISSHOULDNTAFFECTWHATTHEENDUSERDOESTHEDATABASEENGINEWILLTAKECAREOFTHEDATA,ANDTHEFRONTENDWILLMANAGETHEQUERYANDDISPLAYTHEOUTPUTIFTHEFRONTENDSAREWRITTENCORRECTLY,THEDATASHOULDBERELIABLEINANYCASETHEMETHODOFCOMMUNICATIONCOULD,HOWEVER,AFFECTTHESPEEDOFRETRIEVINGDATATHEREAREAFEWIMPORTANTTHINGSTOMENTIONFIRST,NONEOFTHESEFRONTENDSISAVAILABLEYETBUTITSALREADYCLEARTHATTHESEFRONTENDSWILLEXIST,PROBABLYINLARGENUMBERSBYTHEENDOFNESTYEARSECOND,DESPITETHEFACTTHATTHESEFRONTENDSMAYBECONNECTEDTODATABASESERVER,MOSTWILLCONTINUETOINCLUDETHEIROWNLOCALDATABASEENGINES,WHICHWORKONTHEUSERSWORKSTATIONTHATSDATABASE（FORSECURITYORJUSTFORLOGISTICALREASONS）ANDBECAUSENOTALLUSERSWILLBECONNECTEDTOANETWORK。THEMODERNDATADICTIONARYDECISIONMAKINGATALLLEVELSOFMANAGEMENTREQUIRESCOMPLETE,ACCURATE,ANDTIMELYINFORMATION,THEINCREASINGUSEOFCOMPUTERSANDINFORMATIONPROCESSINGTECHNOLOGYINALLENTERPRISESHASHEIGHTENEDMANAGEMENTSAWARENESSOFTHEADVANTAGESOFUTILIZINGDATAASARESOURCEINDEVELOPINGTOOLSANDMETHODOLOGIESTOFACILITATETHEMANAGEMENTOFDATAASARESOURCETHEDATADICTIONARYDDISONESUCHTOOLTHEDDISASOFTWARETOOLTHATPROVIDESFORLOGICALCENTRALIZATIONOFINFORMATIONABOUTANENTERPRISESDATARESOURCESASSUCH,THISTOOLPLAYSAKEYROLEINAIDINGTHEMANAGEMENTANDCONTROLOFTHEENTERPRISESDATARESOURCESTHEDDISUSEFULINALLASPECTSOFANENTERPRISESDATAPROCESSINGENVIRONMENTINCLUDINGDATABASERELATEDAPPLICATIONSANDNONDATABASERELATEDAPPLICATIONADATADICTIONARYISACOMPUTERIZEDCOLLECTIONOFRECORDLAYOUTSTOSUPPORTDATADASEAPPLICATIONS,ITMUSTINCLUDEFIVEFEATURESDATAPROCESSINGDOCUMENTATION,USERDOCUMENTATION,INDEPENDENTDATAELEMENTS,ACCESSIBILITYTOAPPLICATIONLANGUAGESANDACCESSIBILITYTOTHEDATABASEMANAGEMENTSYSTEMRECORDLAYOUTSTELLWHATKINDOFDATAISFOUNDINAFILEAPERSONNELFILESLAYOUT,FOREXAMPLE,SHOWSTHATEACHRECORDHLODSANEMPLOYEENUMBER,NAME,JOBTITLE,SALARYANDTHELIKEFOREACHDATAITEM,THELAYOUTTELLSHOWLONGTHEITEMISNAMEMIGHTBE50LETTERSLONG,JOBTITLEMIGHTBE15,WHETHERAARITHMETICMAYBEPERFORMEDONIT“YES”FORSALARY,”NO”FOREMPLOYEENUMBERAND,IFSO,HOWTHENUMBERISENCODEDBINARY,FLOATINGPOINT,PACKEDDECIMALTHEPERSONNELFILESLAYOUTISABLUEPRINTORTEMPLATETOWHICHEVERYEMPLOYEERECORDCONFORMSTHELAYOUTITSELFTELLSNOTHINGABOUTANYPARTICULAREMPLOYEETHATISTHEROLEOFTHERECORDSDESCRIBEDBYTHELAYOUTINOTHERWORDS,THEPERSONNELFILEHOLDSINFORMATIONABOUTEMPLOYEESCALLIT“PEOPLEDATA,”THELAYOUTHOLDSINFORMATIONABOUTTHEDATAINTHEPERSONNELFILEDATAABOUTDATAINPHILOSOPHICALTERMS,ONECOULDCALLIT“METADATE”DEVELOPERSMUSTKNOWTHESIZE,FORMATANDENCODINGOFAFILESDATAITEMSINORDERTOWRITEPROGRAMSTHATSTOREDATAINTOORRETRIEVEDATAFROMITSO,RECORDLAYOUTSHAVEBEENANESSENTIALTOAPPLICATIONDOCUMENTATIONSINCELONGBEFORETHEDATABASENOTIONBECAMEPOSSIBLEUNTILTHEMID1970S,RECORDLAYOUTCOLLECTIONSWEREOFTENKEPTINTHREERINGBINDERSABSOLUTEPRECISION,HOWEVER,ISRAREINHUMANAFFAIRSMANUALLYMAINTAINEDRECORDLAYOUTSINEVITABLYCONTAINAFEWERRORS,SOMETYPOGRAPHICAL,OTHERSRESULTINGFROMTHECOPYBEINGSLIGHTLYOUTOFDATE“SLIGHTLYOUTOFDATE”FORRECORDLAYOUTS,BYTHEWAY,ISAKINTOBEING“ABITPREGNANT”O(jiān)R“SOMEWHATDEAD”FORSTANDALONEAPPLICATIONS,MINORDOCUMENTATIONFLAWSAREANUISANCEBUTTHEYCANBETOLERATEDBECAUSETHEDEVELOPERMUSTBEINTIMATELYFAMILIARWITHEVERYASPECTOFTHEAPPLICATIONANYWAYTHEREISNOONEELSEDATABASEAPPLICATIONS,ONTHEOTHERHAND,CANNOTASEASLLYTOLERATEEVENMINORERRORSPOTENTIALLY,THESEAPPLICATIONSOFFERTHEPROMISEOFREDUCEDDEVELOPMENTCOSTSWECANTAKEADVANTAGEOFDATAALREADYRECORDEDBYOTHERSTOREAPTHEREWARD,HOWEVER,DEVELOPERSMUSTBEABLETOFINDOUT,WITHLITTLEEFFORT,WHATISINTHEDATABASEEFFORTMEANSCOSTIFFACTFINDINGDEMANDEDDETAILEDEXAMINATIONOFTHEFILES,THEEFFORTWOULDBESPENT,THEPRICEPAID,THEPROMISEBROKENINSHORT,MINORERRORSINMETADATA,ANUISANCETOSTANDALONEDEVELOPMENT,DESTROYONEOFTHEVERYREASONSFORBUILDINGDATABASEAPPLICATIONSTODAYSDICTIONARIESOFFERABSOLUTELYACCURATEESSENTIALMETADATAITEMLENGTH,FORMATANDENCODINGITISTHEIRUNSUNGHISTORYMAKINGACHIEVEMENTFEWEVENTSHAVEAFFECTEDTHEAPPLICATIONBUILDINGINDUSTRYASPROFOUNDLYASPROFOUNDLYASTHESIMPLEFACTTHAT,FORTHEFIRSTTIME,WEHAVERECORDLAYOUTSTHATARETRULY,REALLY,100ACCURATEDATABASEUSERSNEEDPRECISEINFORMATIONABOUTTHEIRDATAINANUTSHELL,DATABASEMEANSYOURACCOUNTSPAYABLEMANAGERAGREESTOUSEVENDORDAARECORDEDBYTHEPURCHASINGDEPARTMENTAGREEMENTHINGESONCONFIDENCE,ANDCONFIDENCEDEPENDSONKNOWINGTHEDETAILSWHOINPURCHASINGISRESPONSIBLEFORITHOWOFTENDOTHEYUPDATEITHOWISITVALIDATEDHOWOFTENDOTHEYMAKEMISTAKESIFISEEANERROR,MAYICORRECTITORMUSTIASKPURCHASINGTODOSOTHEQUESTIONSAREASHARDTOPREDICTASTHEYAREIMPORTANTTHEANSWERSSHOULDBERECORDEDINTHEDICTIONARYTHESECONDFUNCTIONOFADATADICTIONARY,THEN,ISTOSTOREFACTSABOUTTHEDATABASETHAT,ALTHOUGHIRRELEVANTTODP,AREVITALTOOURUSERS,THECHALLENGETODICTIONARYMAKERSISTHAT,INCONTRASTTODPMETADATA,YOUREALLYDONTKNOWWHICHQUESTIONSUSERSCONSIDERIMPORTANTUNTILTHEYASKTHEMHENCE,ADICTIONARYSHOULDLETYOUDEFINENEWCLASSESOFDATAABOUTDATAADICTIONARYSTHIRDFUNCTIONISTOENABLEUSTORECORD,INONEPLACE,FACTSABOUTADATAELEMENTTHATDONOTVARYFROMONERECORDLAYOUTTOTHENEXTEMPLOYEENUMBERS,SAY,MAYAPPEARINTHEDATABASESPERSONNELRECORDASITSKEY,INTHEDEPARTMENTRECORDIDENTIFYINGTHEUNITMANAGER,INTHEMAINTENANCEWORKORDERRECORDTHEENGINEERRESPONSIBLEANDINMANYOTHERS。NEVERTHELESS,VOMMONSENSEDICTATESTHAT,WHEREVERITISUSED,INWHATEVERPROGRAMITAPPEARS,THEEMPLOYEENUMBERMUSTAPPEARAS“EMPLOYEENUM,”NINECHARACTERSLONGANDNOTENCODEDFORMATHWEPUTVENDORADDRESSINVENDORMASTERRECORDS,NOTINEVERYPURCHASEORDERSIMILARLY,DATAABOUTTHEABSTRACTION“EMPLOYEENUMBER”SHOULDBERECORDEDINAFILEOFDATAELEMENTDEFINITIONS,NOTINEVERYRECORDLAYOUTINWHICHITAPPEARS。中文翻譯分布式數(shù)據(jù)庫(kù)的發(fā)展從七十年代初開(kāi)始，集中式數(shù)據(jù)庫(kù)已在數(shù)據(jù)處理產(chǎn)業(yè)中大規(guī)模地使用，分布式數(shù)據(jù)庫(kù)比集中式數(shù)據(jù)庫(kù)要復(fù)雜得多，它們涉及復(fù)雜的軟件來(lái)處理諸如并發(fā)控制、死鎖保護(hù)和在多點(diǎn)更新中保持?jǐn)?shù)據(jù)的完整性等。只是到了最近，這些軟件問(wèn)題才被認(rèn)識(shí)到，配有這種軟件的產(chǎn)品也才剛剛開(kāi)始在市場(chǎng)上出現(xiàn)。當(dāng)代的分布式數(shù)據(jù)庫(kù)主要是重復(fù)多個(gè)點(diǎn)上的數(shù)據(jù)庫(kù)而不是分布單個(gè)數(shù)據(jù)庫(kù)。為了把數(shù)據(jù)放到盡量靠近作用率最高的地方，下一代數(shù)據(jù)庫(kù)可能涉及把數(shù)據(jù)分布在幾個(gè)不同地理位置的節(jié)點(diǎn)上。利用這些數(shù)據(jù)庫(kù)，多點(diǎn)更新可能比現(xiàn)在更經(jīng)常地進(jìn)行。為了確保數(shù)據(jù)庫(kù)完整性被破壞的風(fēng)險(xiǎn)降至絕對(duì)小，這就要非常小心地監(jiān)視網(wǎng)絡(luò)和分布數(shù)據(jù)庫(kù)，以便使操作人員對(duì)發(fā)生數(shù)據(jù)破壞的任何形勢(shì)作出快速反應(yīng)。在這方面已經(jīng)做了一些工作，利用專家系統(tǒng)有助于確定某個(gè)問(wèn)題是通信問(wèn)題還是數(shù)據(jù)庫(kù)問(wèn)題。然而還需使這些診斷技術(shù)進(jìn)一步成熟到使計(jì)算機(jī)操作人員能天天使用它們的程度。在不遠(yuǎn)的將來(lái)，最大量的開(kāi)發(fā)工作將在優(yōu)化器領(lǐng)域中進(jìn)行。它是一個(gè)由分布式數(shù)據(jù)庫(kù)采用的模塊，以便將計(jì)算各通信成本降至最小其中強(qiáng)調(diào)通信成本，因與處理成本相比本通常更高。先進(jìn)的優(yōu)化器將考慮通信成本、磁盤I/O和處理成本，還半評(píng)價(jià)幾種在網(wǎng)絡(luò)上傳送數(shù)據(jù)的方法，以滿足多交易處理的要求。它也將評(píng)價(jià)幾種對(duì)數(shù)據(jù)庫(kù)的存取路徑。在某些情況下它還將存儲(chǔ)關(guān)于不同存取策略比較的信息，以供將來(lái)使用。有些優(yōu)化器還將打印出它們所做過(guò)事情的細(xì)節(jié)。作為一般準(zhǔn)則，當(dāng)分布數(shù)據(jù)庫(kù)中的節(jié)點(diǎn)數(shù)超過(guò)五個(gè)時(shí)，優(yōu)化問(wèn)題就變得復(fù)雜了。另一個(gè)重要的未來(lái)發(fā)展可能在系統(tǒng)體系結(jié)構(gòu)的領(lǐng)域中發(fā)生。最近研究的大量設(shè)想鼓吹能給分布式數(shù)據(jù)庫(kù)作基礎(chǔ)的最吸引人的模式是客戶服務(wù)器模式。在這種模式中，“客戶”（數(shù)據(jù)用戶）與“服務(wù)器”（數(shù)據(jù)提供者）是分開(kāi)的，通過(guò)高速局域網(wǎng)將它們連接起來(lái)。這個(gè)模式的吸引人處是服務(wù)器能以模塊形式很容易地加進(jìn)去，在一定程度上能將不同類型的硬件混接在一起。然而，這與今后局域網(wǎng)技術(shù)更復(fù)雜的發(fā)展休戚相關(guān)，特別是要與支持異種硬件結(jié)合。目前能反映出客戶服務(wù)器模式的最著名產(chǎn)品是SYBASE、INGRES和TANDEM公司的NONSTOPSQL。目前，分布式數(shù)據(jù)庫(kù)是多家硬件和DBMS供應(yīng)商十分感興趣的題目。假設(shè)諸如恢復(fù)、診斷、優(yōu)化和局域網(wǎng)結(jié)構(gòu)中的問(wèn)題能得到解決，分布數(shù)據(jù)庫(kù)看來(lái)是九十年代計(jì)算機(jī)工業(yè)發(fā)展的一個(gè)重要領(lǐng)域。數(shù)據(jù)庫(kù)分為中央服務(wù)器和本地前端我們已聽(tīng)到很多關(guān)于數(shù)據(jù)庫(kù)市場(chǎng)上發(fā)生的變化，也許太多了。你知道了SQL、數(shù)據(jù)庫(kù)服務(wù)器、客戶/服務(wù)器關(guān)系等經(jīng)常出現(xiàn)的術(shù)語(yǔ)。讓我們以個(gè)人計(jì)算機(jī)數(shù)據(jù)庫(kù)基本發(fā)展為開(kāi)端。個(gè)人機(jī)上的數(shù)據(jù)庫(kù)一般分為兩大類。平文件管理程序一次只與單個(gè)文件打交道，通常是菜單驅(qū)動(dòng)，易用，供最終用戶維持其自己的數(shù)據(jù)（如電話或地址表）管理。關(guān)系數(shù)據(jù)庫(kù)一交可與多個(gè)文件打交道，它們主要供顧問(wèn)或公司內(nèi)編制應(yīng)用程序的支持人員使用。在過(guò)去幾年內(nèi)，這些區(qū)別開(kāi)始被打破。一方面，平文件管理程序變得更強(qiáng)，在基本功能方面開(kāi)始向關(guān)系數(shù)據(jù)庫(kù)靠攏。關(guān)系數(shù)據(jù)庫(kù)也在。PARADOX一類的程序設(shè)計(jì)對(duì)執(zhí)行特定詢問(wèn)的用戶很簡(jiǎn)單。DATAEASE一類的產(chǎn)品讓你很少或者不用編程就能設(shè)計(jì)多文件應(yīng)用程序。同時(shí)，其它的數(shù)據(jù)庫(kù)產(chǎn)品強(qiáng)調(diào)能力和多用戶的特點(diǎn)超過(guò)過(guò)了易用性。這對(duì)當(dāng)初為小型機(jī)和大型機(jī)設(shè)計(jì)，現(xiàn)在卻由PC版提供各支持的功能真正強(qiáng)大的數(shù)據(jù)庫(kù)驅(qū)動(dòng)系統(tǒng)來(lái)說(shuō)更是如此。其中最成功的例子是ORACLE，它是一個(gè)負(fù)有成名的強(qiáng)大數(shù)據(jù)庫(kù)驅(qū)動(dòng)系統(tǒng)，設(shè)計(jì)用于多用戶環(huán)境，但缺乏多數(shù)PC用戶期待的易用性。占據(jù)市場(chǎng)首位的DBASE也的試圖變化，以滿足各種期望。DBASEIV增加了一個(gè)前端，使之更易使用、供特定詢問(wèn)的按例詢問(wèn)接口以及以便晚高級(jí)的索引和關(guān)系的形式變得更強(qiáng)大。所有這些是出現(xiàn)在這樣的時(shí)刻，多數(shù)數(shù)據(jù)庫(kù)公司認(rèn)識(shí)到，我們正在步入這樣一個(gè)世界，即很多數(shù)據(jù)庫(kù)系統(tǒng)的商業(yè)用戶與局域網(wǎng)或者通過(guò)網(wǎng)關(guān)與小型機(jī)各大型機(jī)相連?？紤]一下數(shù)據(jù)庫(kù)是如何發(fā)展過(guò)來(lái)的。當(dāng)我們只有單用戶時(shí)，數(shù)據(jù)庫(kù)是建在本地硬盤上。隨著我們朝多用戶數(shù)據(jù)庫(kù)進(jìn)展時(shí)，接下來(lái)的步伐是使數(shù)據(jù)庫(kù)本身留駐在中央文件服務(wù)器上，現(xiàn)在所有主要的PC數(shù)據(jù)庫(kù)都可以這樣。雖然涉及的網(wǎng)絡(luò)可以是基于服務(wù)器的（如NOVELL的NETWARE）或?qū)Φ鹊模ㄈ?0NET），但信息只存在網(wǎng)絡(luò)磁盤上，本地工作站或個(gè)人機(jī)完成所有處理。今天帶有LANPACK的DBASEPLUS或、RBASE和PARADOX就是這種情況。所有這一切并不太復(fù)雜，但它能啊作。然而現(xiàn)在，隨著我們朝著更強(qiáng)的PC機(jī)和更復(fù)雜的網(wǎng)絡(luò)操作系統(tǒng)發(fā)展，各公司開(kāi)始采取新的方法。一個(gè)很大的差別是現(xiàn)在各公司把數(shù)據(jù)庫(kù)看成兩部分前端和數(shù)據(jù)庫(kù)驅(qū)動(dòng)系統(tǒng)。前端控制數(shù)據(jù)如何在屏幕上顯示，用戶如何與數(shù)據(jù)打交道；數(shù)據(jù)庫(kù)驅(qū)動(dòng)系統(tǒng)實(shí)際上保留數(shù)據(jù)和執(zhí)行諸如查詢、排序和索引等操作。這就是常常所說(shuō)的客戶/服務(wù)器方式，以各臺(tái)個(gè)人機(jī)作為集中式數(shù)據(jù)庫(kù)服務(wù)器的客戶?！翱蛻?服務(wù)器”意味著這個(gè)產(chǎn)品的運(yùn)行在本質(zhì)上是與網(wǎng)絡(luò)相連的。SQL服務(wù)器是軟件數(shù)據(jù)庫(kù)“驅(qū)動(dòng)系統(tǒng)”，它可以在網(wǎng)絡(luò)服務(wù)器、專用網(wǎng)絡(luò)工作站或者運(yùn)行OS/2的非專用工作站上工作。這個(gè)機(jī)器完成所謂的“后端處理功能”，其中包括管理交易、安全、排序和讀寫數(shù)據(jù)?？蛻?服務(wù)器結(jié)構(gòu)背后的思想是把處理負(fù)荷分散到適合它的地方。例如，服務(wù)器承擔(dān)處理適用于每個(gè)助記詞的任務(wù)（安全、存取共同數(shù)據(jù)等）。前端應(yīng)用程序?qū)τ脩艚涌趤?lái)說(shuō)可合性能最佳。這些應(yīng)用的處理是在需要據(jù)處當(dāng)?shù)剡M(jìn)行。這樣，無(wú)用的網(wǎng)絡(luò)信息流量減少了。這就是說(shuō)隨著使用SQL服務(wù)器的人數(shù)增加，造成的網(wǎng)絡(luò)性能下降較小。一家市場(chǎng)研究公司說(shuō)“隨著符合這種新的客戶/服務(wù)器結(jié)構(gòu)產(chǎn)品的推出，供應(yīng)商們將爭(zhēng)奪市場(chǎng)份額，而用戶可能旁觀最初涌現(xiàn)的產(chǎn)品。”IBM推出了OS/2擴(kuò)展版，它給OS/2增加了IBM自己的數(shù)據(jù)管理程序，本質(zhì)上它是流行的大型機(jī)數(shù)據(jù)庫(kù)DB2的微型機(jī)版本。ASHTONTATE公司各MICROSOFT公司開(kāi)始交付它們的數(shù)據(jù)庫(kù)服務(wù)器SQLSERVER,這是以SYBASE公司一種數(shù)據(jù)庫(kù)驅(qū)動(dòng)系統(tǒng)為基礎(chǔ)的。這兩具產(chǎn)品能以幾種方法被尋址用一種語(yǔ)言（如C）寫程序，采用SQL發(fā)送數(shù)據(jù)庫(kù)查詢，或者使用另一個(gè)起前端作用的數(shù)據(jù)庫(kù)和程序。就它們本身來(lái)說(shuō)，這些軟件包不是作為前端或者另一種數(shù)據(jù)庫(kù)編程語(yǔ)言設(shè)計(jì)的。ORACLE公司已開(kāi)始使其數(shù)據(jù)庫(kù)按同一方法發(fā)展，以作為一個(gè)由多種前端訪問(wèn)的數(shù)據(jù)庫(kù)驅(qū)動(dòng)系統(tǒng)。ORACLE已經(jīng)有可供123、DBXL和MYPERCARD使用的接口。其他起始于小型機(jī)世界的數(shù)據(jù)庫(kù)公司都采取類似的方法。MICRORIM公司已提議一個(gè)變種，即代號(hào)為STLAS的新產(chǎn)品系列，它被設(shè)計(jì)成能管理和協(xié)調(diào)以多種其它格式存放的數(shù)據(jù)、其它數(shù)據(jù)庫(kù)服務(wù)器保持的數(shù)據(jù)和包括RBASE、DBASE和OS/2擴(kuò)展版以及ORACLE、INGRES、INFORMIX和UNIFY等多數(shù)小型機(jī)數(shù)據(jù)庫(kù)中的數(shù)據(jù)。ATLAS驅(qū)動(dòng)系統(tǒng)也被設(shè)計(jì)成能通過(guò)多個(gè)前端進(jìn)行訪問(wèn)（MICRORIM將前端稱作”表面”）。在這樣的環(huán)境中，數(shù)據(jù)庫(kù)驅(qū)動(dòng)系統(tǒng)（可以是OS/2擴(kuò)展版、SQLSERVER、ORACLE、GUPTA公司的SQLBASE或者完全不現(xiàn)的其它產(chǎn)品）留駐在中央數(shù)據(jù)庫(kù)服務(wù)器，通常是專用于此目的的計(jì)算機(jī)和硬磁盤。各個(gè)最終用戶就能使用各種不同的前端或者用各種其它數(shù)據(jù)庫(kù)語(yǔ)言開(kāi)發(fā)的應(yīng)用程序。各個(gè)前端如何與各個(gè)數(shù)據(jù)庫(kù)驅(qū)動(dòng)系統(tǒng)通信是有差別的。多數(shù)產(chǎn)品采用SQL的某個(gè)變型，SQL是通過(guò)IBM的DB2流行起來(lái)的數(shù)據(jù)庫(kù)查詢語(yǔ)言。有些產(chǎn)品將發(fā)送實(shí)際的SQL查詢，有些將通過(guò)API（應(yīng)用程序接口）進(jìn)行通信，有些用服務(wù)器自有的程序發(fā)送信息。但這不應(yīng)影響最終用戶所做的事數(shù)據(jù)庫(kù)驅(qū)動(dòng)系統(tǒng)看管數(shù)據(jù)，前端管理查詢和顯示輸出。如果前端寫的正確，數(shù)據(jù)在任何情況都應(yīng)可靠。然而，通信方法可能影響檢索數(shù)據(jù)的速度。有些重要的事情要提醒。首先，這些前端還一個(gè)都沒(méi)有上市。但很清楚，這些前端將會(huì)存在，可能在明年年底大量出現(xiàn)。其次，盡管有這些前端可以與數(shù)據(jù)庫(kù)服務(wù)器相連這樣一個(gè)事實(shí)，但多數(shù)前端將包含它們自己的本地?cái)?shù)據(jù)庫(kù)驅(qū)動(dòng)系統(tǒng)。它們?cè)谟脩艄ぷ髡旧瞎ぷ?。這是因?yàn)槟悴恍枰阉械男畔?chǔ)存在共享數(shù)據(jù)庫(kù)中（出于安全或只是出于合理），并且由于不是所有的用戶都將與網(wǎng)絡(luò)相連。現(xiàn)代數(shù)據(jù)字典在各級(jí)管理人員作決策時(shí)都需要完整