1．Databasemanagementsystem

1.Databasemanagementsystem

ADatabaseManagementSystem(DBMS)isasetofcomputerprogramsthatcontrolsthecreation,maintenance,andtheuseofadatabase.Itallowsorganizationstoplacecontrolofdatabasedevelopmentinthehandsofdatabaseadministrators(DBAs)andotherspecialists.ADBMSisasystemsoftwarepackagethathelpstheuseofintegratedcollectionofdatarecordsandfilesknownasdatabases.Itallowsdifferentuserapplicationprogramstoeasilyaccessthesamedatabase.DBMSsmayuseanyofavarietyofdatabasemodels,suchasthenetworkmodelorrelationalmodel.Inlargesystems,aDBMSallowsusersandothersoftwaretostoreandretrievedatainastructuredway.Insteadofhavingtowritecomputerprogramstoextractinformation,usercanasksimplequestionsinaquerylanguage.Thus,manyDBMSpackagesprovideFourth-generationprogramminglanguage(4GLs)andotherapplicationdevelopmentfeatures.Ithelpstospecifythelogicalorganizationforadatabaseandaccessandusetheinformationwithinadatabase.Itprovidesfacilitiesforcontrollingdataaccess,enforcingdataintegrity,managingconcurrency,andrestoringthedatabasefrombackups.ADBMSalsoprovidestheabilitytologicallypresentdatabaseinformationtousers.

2.Overview

ADBMSisasetofsoftwareprogramsthatcontrolstheorganization,storage,management,andretrievalofdatainadatabase.DBMSsarecategorizedaccordingtotheirdatastructuresortypes.TheDBMSacceptsrequestsfordatafromanapplicationprogramandinstructstheoperatingsystemtotransfertheappropriatedata.Thequeriesandresponsesmustbesubmittedandreceivedaccordingtoaformatthatconformstooneormoreapplicableprotocols.WhenaDBMSisused,informationsystemscanbechangedmuchmoreeasilyastheorganization'sinformationrequirementschange.Newcategoriesofdatacanbeaddedtothedatabasewithoutdisruptiontotheexistingsystem.

DatabaseserversarecomputersthatholdtheactualdatabasesandrunonlytheDBMSandrelatedsoftware.Databaseserversareusuallymultiprocessorcomputers,withgenerousmemoryandRAIDdiskarraysusedforstablestorage.Hardwaredatabaseaccelerators,connectedtooneormoreserversviaahigh-speedchannel,arealsousedinlargevolumetransactionprocessingenvironments.DBMSsarefoundattheheartofmostdatabaseapplications.DBMSsmaybebuiltaroundacustommultitaskingkernelwithbuilt-innetworkingsupport,butmodernDBMSstypicallyrelyonastandardoperatingsystemtoprovidethesefunctions.

3.History

Databaseshavebeeninusesincetheearliestdaysofelectroniccomputing.Unlikemodernsystemswhichcanbeappliedtowidelydifferentdatabasesandneeds,thevastmajorityofoldersystemsweretightlylinkedtothecustomdatabasesinordertogainspeedattheexpenseofflexibility.OriginallyDBMSswerefoundonlyinlargeorganizationswiththecomputerhardwareneededtosupportlargedatasets.

3.11960sNavigationalDBMS

Ascomputersgrewinspeedandcapability,anumberofgeneral-purposedatabasesystemsemerged;bythemid-1960stherewereanumberofsuchsystemsincommercialuse.Interestinastandardbegantogrow,andCharlesBachman,authorofonesuchproduct,IntegratedDataStore(IDS),foundedthe"DatabaseTaskGroup"withinCODASYL,thegroupresponsibleforthecreationandstandardizationofCOBOL.In1971theydeliveredtheirstandard,whichgenerallybecameknownasthe"Codasylapproach",andsoontherewereanumberofcommercialproductsbasedonitavailable.

TheCodasylapproachwasbasedonthe"manual"navigationofalinkeddatasetwhichwasformedintoalargenetwork.Whenthedatabasewasfirstopened,theprogramwashandedbackalinktothefirstrecordinthedatabase,whichalsocontainedpointerstootherpiecesofdata.Tofindanyparticularrecordtheprogrammerhadtostepthroughthesepointersoneatatimeuntiltherequiredrecordwasreturned.Simplequerieslike"findallthepeopleinIndia"requiredtheprogramtowalktheentiredatasetandcollectthematchingresults.Therewas,essentially,noconceptof"find"or"search".Thismightsoundlikeaseriouslimitationtoday,butinanerawhenthedatawasmostoftenstoredonmagnetictapesuchoperationsweretooexpensivetocontemplateanyway.

IBMalsohadtheirownDBMSsystemin1968,knownasIMS.IMSwasadevelopmentofsoftwarewrittenfortheApolloprogramontheSystem/360.IMSwasgenerallysimilarinconcepttoCodasyl,butusedastricthierarchyforitsmodelofdatanavigationinsteadofCodasyl'snetworkmodel.Bothconceptslaterbecameknownasnavigationaldatabasesduetothewaydatawasaccessed,andBachman's1973TuringAwardawardpresentationwasTheProgrammerasNavigator.IMSisclassifiedasahierarchicaldatabase.IMSandIDMS,bothCODASYLdatabases,aswellasCINCOMsTOTALdatabaseareclassifiedasnetworkdatabases.

3.21970sRelationalDBMS

EdgarCoddworkedatIBMinSanJose,California,inoneoftheiroffshootofficesthatwasprimarilyinvolvedinthedevelopmentofharddisksystems.HewasunhappywiththenavigationalmodeloftheCodasylapproach,notablythelackofa"search"facilitywhichwasbecomingincreasinglyuseful.In1970,hewroteanumberofpapersthatoutlinedanewapproachtodatabaseconstructionthateventuallyculminatedinthegroundbreakingARelationalModelofDataforLargeSharedDataBanks.[1]

Inthispaper,hedescribedanewsystemforstoringandworkingwithlargedatabases.Insteadofrecordsbeingstoredinsomesortoflinkedlistoffree-formrecordsasinCodasyl,Codd'sideawastousea"table"offixed-lengthrecords.Alinked-listsystemwouldbeveryinefficientwhenstoring"sparse"databaseswheresomeofthedataforanyonerecordcouldbeleftempty.Therelationalmodelsolvedthisbysplittingthedataintoaseriesofnormalizedtables,withoptionalelementsbeingmovedoutofthemaintabletowheretheywouldtakeuproomonlyifneeded.

Forinstance,acommonuseofadatabasesystemistotrackinformationaboutusers,theirname,logininformation,variousaddressesandphonenumbers.Inthenavigationalapproachallofthesedatawouldbeplacedinasinglerecord,andunuseditemswouldsimplynotbeplacedinthedatabase.Intherelationalapproach,thedatawouldbenormalizedintoausertable,anaddresstableandaphonenumbertable(forinstance).Recordswouldbecreatedintheseoptionaltablesonlyiftheaddressorphonenumberswereactuallyprovided.

Linkingtheinformationbacktogetheristhekeytothissystem.Intherelationalmodel,somebitofinformationwasusedasa"key",uniquelydefiningaparticularrecord.Wheninformationwasbeingcollectedaboutauser,informationstoredintheoptional(orrelated)tableswouldbefoundbysearchingforthiskey.Forinstance,iftheloginnameofauserisunique,addressesandphonenumbersforthatuserwouldberecordedwiththeloginnameasitskey.This"re-linking"ofrelateddatabackintoasinglecollectionissomethingthattraditionalcomputerlanguagesarenotdesignedfor.

Justasthenavigationalapproachwouldrequireprogramstoloopinordertocollectrecords,therelationalapproachwouldrequireloopstocollectinformationaboutanyonerecord.Codd'ssolutiontothenecessaryloopingwasaset-orientedlanguage,asuggestionthatwouldlaterspawntheubiquitousSQL.Usingabranchofmathematicsknownastuplecalculus,hedemonstratedthatsuchasystemcouldsupportalltheoperationsofnormaldatabases(inserting,updatingetc.)aswellasprovidingasimplesystemforfindingandreturningsetsofdatainasingleoperation.

Codd'spaperwaspickedupbytwopeopleattheBerkeley,EugeneWongandMichaelStonebraker.TheystartedaprojectknownasINGRESusingfundingthathadalreadybeenallocatedforageographicaldatabaseproject,usingstudentprogrammerstoproducecode.Beginningin1973,INGRESdelivereditsfirsttestproductswhichweregenerallyreadyforwidespreadusein1979.Duringthistime,anumberofpeoplehadmoved"through"thegroup—perhapsasmanyas30peopleworkedontheproject,aboutfiveatatime.INGRESwassimilartoSystemRinanumberofways,includingtheuseofa"language"fordataaccess,knownasQUEL—QUELwasinfactrelational,havingbeenbasedonCodd'sownAlphalanguage,buthassincebeencorruptedtofollowSQL,thusviolatingmuchthesameconceptsoftherelationalmodelasSQLitself.

IBMitselfdidonetestimplementationoftherelationalmodel,PRTV,andaproductionone,BusinessSystem12,bothnowdiscontinued.HoneywelldidMRDSforMultics,andnowtherearetwonewimplementations:AlphoraDataphorandRel.AllotherDBMSimplementationsusuallycalledrelationalareactuallySQLDBMSs.In1968,theUniversityofMichiganbegandevelopmentoftheMicroDBMSrelationaldatabasemanagementsystem.ItwasusedtomanageverylargedatasetsbytheUSDepartmentofLabor,theEnvironmentalProtectionAgencyandresearchersfromUniversityofAlberta,theUniversityofMichiganandWayneStateUniversity.ItranonmainframecomputersusingMichiganTerminalSystem.Thesystemremainedinproductionuntil1996.

3.3End1970sSQLDBMS

IBMstartedworkingonaprototypesystemlooselybasedonCodd'sconceptsasSystemRintheearly1970s.Thefirstversionwasreadyin1974/5,andworkthenstartedonmulti-tablesystemsinwhichthedatacouldbesplitsothatallofthedataforarecord(muchofwhichisoftenoptional)didnothavetobestoredinasinglelarge"chunk".Subsequentmulti-userversionsweretestedbycustomersin1978and1979,bywhichtimeastandardizedquerylanguage,SQL,hadbeenadded.Codd'sideaswereestablishingthemselvesasbothworkableandsuperiortoCodasyl,pushingIBMtodevelopatrueproductionversionofSystemR,knownasSQL/DS,and,later,Database2(DB2).

ManyofthepeopleinvolvedwithINGRESbecameconvincedofthefuturecommercialsuccessofsuchsystems,andformedtheirowncompaniestocommercializetheworkbutwithanSQLinterface.Sybase,Informix,NonStopSQLandeventuallyIngresitselfwereallbeingsoldasoffshootstotheoriginalINGRESproductinthe1980s.EvenMicrosoftSQLServerisactuallyare-builtversionofSybase,andthus,INGRES.OnlyLarryEllison'sOraclestartedfromadifferentchain,basedonIBM'spapersonSystemR,andbeatIBMtomarketwhenthefirstversionwasreleasedin1978.

StonebrakerwentontoapplythelessonsfromINGREStodevelopanewdatabase,Postgres,whichisnowknownasPostgreSQL.PostgreSQLisoftenusedforglobalmissioncriticalapplications(domainnameregistriesuseitastheirprimarydatastore,asdomanylargecompaniesandfinancialinstitutions).

InSweden,Codd'spaperwasalsoreadandMimerSQLwasdevelopedfromthemid-70satUppsalaUniversity.In1984,thisprojectwasconsolidatedintoanindependententerprise.Intheearly1980s,Mimerintroducedtransactionhandlingforhighrobustnessinapplications,anideathatwassubsequentlyimplementedonmostotherDBMS.

3.41980sObjectOrientedDatabases

The1980s,alongwithariseinobjectorientedprogramming,sawagrowthinhowdatainvariousdatabaseswerehandled.Programmersanddesignersbegantotreatthedataintheirdatabasesasobjects.Thatistosaythatifaperson'sdatawereinadatabase,thatperson'sattributes,suchastheiraddress,phonenumber,andage,werenowconsideredtobelongtothatpersoninsteadofbeingextraneousdata.Thisallowsforrelationshipsbetweendatatoberelationtoobjectsandtheirattributesandnottoindividualfields.

Anotherbiggamechangerfordatabasesinthe1980swasthefocusonincreasingreliabilityandaccessspeeds.In1989,twoprofessorsfromtheUniversityofMichiganatMadison,publishedanarticleatanACMassociatedconferenceoutliningtheirmethodsonincreasingdatabaseperformance.Theideawastoreplicatespecificimportant,andoftenqueriedinformation,andstoreitinasmallertemporarydatabasethatlinkedthesekeyfeaturesbacktothemaindatabase.Thismeantthataquerycouldsearchthesmallerdatabasemuchquicker,ratherthansearchtheentiredataset.Thiseventuallyleadstothepracticeofindexing,whichisusedbyalmosteveryoperatingsystemfromWindowstothesystemthatoperatesAppleiPoddevices.

4.DBMSbuildingblocks

ADBMSincludesfourmainparts:modelinglanguage,datastructure,databasequerylanguage,andtransactionmechanisms:

4.1ComponentsofDBMS

DBMSEngineacceptslogicalrequestfromthevariousotherDBMSsubsystems,convertsthemintophysicalequivalents,andactuallyaccessesthedatabaseanddatadictionaryastheyexistonastoragedevice.

DataDefinitionSubsystemhelpsusertocreateandmaintainthedatadictionaryanddefinethestructureofthefilesinadatabase.

DataManipulationSubsystemhelpsusertoadd,change,anddeleteinformationinadatabaseandqueryitforvaluableinformation.Softwaretoolswithinthedatamanipulationsubsystemaremostoftentheprimaryinterfacebetweenuserandtheinformationcontainedinadatabase.Itallowsusertospecifyitslogicalinformationrequirements.

ApplicationGenerationSubsystemcontainsfacilitiestohelpuserstodeveloptransaction-intensiveapplications.Itusuallyrequiresthatuserperformadetailedseriesoftaskstoprocessatransaction.Itfacilitateseasy-to-usedataentryscreens,programminglanguages,andinterfaces.

DataAdministrationSubsystemhelpsuserstomanagetheoveralldatabaseenvironmentbyprovidingfacilitiesforbackupandrecovery,securitymanagement,queryoptimization,concurrencycontrol,andchangemanagement.

4.2Modelinglanguage

AdatamodelinglanguagetodefinetheschemaofeachdatabasehostedintheDBMS,accordingtotheDBMSdatabasemodel.Thefourmostcommontypesofmodelsarethe:

? hierarchicalmodel,

? networkmodel,

? relationalmodel,and

? objectmodel.

Invertedlistsandothermethodsarealsoused.Agivendatabasemanagementsystemmayprovideoneormoreofthefourmodels.Theoptimalstructuredependsonthenaturalorganizationoftheapplication'sdata,andontheapplication'srequirements(whichincludetransactionrate(speed),reliability,maintainability,scalability,andcost).

ThedominantmodelinusetodayistheadhoconeembeddedinSQL,despitetheobjectionsofpuristswhobelievethismodelisacorruptionoftherelationalmodel,sinceitviolatesseveralofitsfundamentalprinciplesforthesakeofpracticalityandperformance.ManyDBMSsalsosupporttheOpenDatabaseConnectivityAPIthatsupportsastandardwayforprogrammerstoaccesstheDBMS.

Beforethedatabasemanagementapproach,organizationsreliedonfileprocessingsystemstoorganize,store,andprocessdatafiles.Endusersbecameaggravatedwithfileprocessingbecausedataisstoredinmanydifferentfilesandeachorganizedinadifferentway.Eachfilewasspecializedtobeusedwithaspecificapplication.Needlesstosay,fileprocessingwasbulky,costlyandnonflexiblewhenitcametosupplyingneededdataaccuratelyandpromptly.Dataredundancyisanissuewiththefileprocessingsystembecausetheindependentdatafilesproduceduplicatedatasowhenupdateswereneededeachseparatefilewouldneedtobeupdated.Anotherissueisthelackofdataintegration.Thedataisdependentonotherdatatoorganizeandstoreit.Lastly,therewasnotanyconsistencyorstandardizationofthedatainafileprocessingsystemwhichmakesmaintenancedifficult.Forallthesereasons,thedatabasemanagementapproachwasproduced.Databasemanagementsystems(DBMS)aredesignedtouseoneoffivedatabasestructurestoprovidesimplisticaccesstoinformationstoredindatabases.Thefivedatabasestructuresarehierarchical,network,relational,multidimensionalandobject-orientedmodels.

ThehierarchicalstructurewasusedinearlymainframeDBMS.Records’relationshipsformatreelikemodel.Thisstructureissimplebutnonflexiblebecausetherelationshipisconfinedtoaone-to-manyrelationship.IBM’sIMSsystemandtheRDMMobileareexamplesofahierarchicaldatabasesystemwithmultiplehierarchiesoverthesamedata.RDMMobileisanewlydesignedembeddeddatabaseforamobilecomputersystem.Thehierarchicalstructureisusedprimarytodayforstoringgeographicinformationandfilesystems.

Thenetworkstructureconsistsofmorecomplexrelationships.Unlikethehierarchicalstructure,itcanrelatetomanyrecordsandaccessesthembyfollowingoneofseveralpaths.Inotherwords,thisstructureallowsformany-to-manyrelationships.

Therelationalstructureisthemostcommonlyusedtoday.Itisusedbymainframe,midrangeandmicrocomputersystems.Itusestwo-dimensionalrowsandcolumnstostoredata.Thetablesofrecordscanbeconnectedbycommonkeyvalues.WhileworkingforIBM,E.F.Codddesignedthisstructurein1970.Themodelisnoteasyfortheendusertorunquerieswithbecauseitmayrequireacomplexcombinationofmanytables.

Themultidimensionalstructureissimilartotherelationalmodel.Thedimensionsofthecubelookingmodelhavedatarelatingtoelementsineachcell.Thisstructuregivesaspreadsheetlikeviewofdata.Thisstructureiseasytomaintainbecauserecordsarestoredasfundamentalattributes,thesamewaythey’reviewedandthestructureiseasytounderstand.Itshighperformancehasmadeitthemostpopulardatabasestructurewhenitcomestoenablingonlineanalyticalprocessing(OLAP).

Theobjectorientedstructurehastheabilitytohandlegraphics,pictures,voiceandtext,typesofdata,withoutdifficultlyunliketheotherdatabasestructures.ThisstructureispopularformultimediaWeb-basedapplications.Itwasdesignedtoworkwithobject-orientedprogramminglanguagessuchasJava.

4.3Datastructure

Datastructures(fields,records,filesandobjects)optimizedtodealwithverylargeamountsofdatastoredonapermanentdatastoragedevice(whichimpliesrelativelyslowaccesscomparedtovolatilemainmemory).

4.4Databasequerylanguage

Adatabasequerylanguageandreportwriterallowsuserstointeractivelyinterrogatethedatabase,analyzeitsdataandupdateitaccordingtotheusersprivilegesondata.Italsocontrolsthesecurityofthedatabase.Datasecuritypreventsunauthorizedusersfromviewingorupdatingthedatabase.Usingpasswords,usersareallowedaccesstotheentiredatabaseorsubsetsofitcalledsubschemas.Forexample,anemployeedatabasecancontainallthedataaboutanindividualemployee,butonegroupofusersmaybeauthorizedtoviewonlypayrolldata,whileothersareallowedaccesstoonlyworkhistoryandmedicaldata.

IftheDBMSprovidesawaytointeractivelyenterandupdatethedatabase,aswellasinterrogateit,thiscapabilityallowsformanagingpersonaldatabases.However,itmaynotleaveanaudittrailofactionsorprovidethekindsofcontrolsnecessaryinamulti-userorganization.Thesecontrolsareonlyavailablewhenasetofapplicationprogramsarecustomizedforeachdataentryandupdatingfunction.

4.5Transactionmechanism

AdatabasetransactionmechanismideallyguaranteesACIDpropertiesinordertoensuredataintegritydespiteconcurrentuseraccesses(concurrencycontrol),andfaults(faulttolerance).Italsomaintainstheintegrityofthedatainthedatabase.TheDBMScanmaintaintheintegrityofthedatabasebynotallowingmorethanoneusertoupdatethesamerecordatthesametime.TheDBMScanhelppreventduplicaterecordsviauniqueindexconstraints;forexample,notwocustomerswiththesamecustomernumbers(keyfields)canbeenteredintothedatabase.SeeACIDpropertiesformoreinformation(Redundancyavoidance).

5.DBMStopics

5.1External,LogicalandInternalview

Adatabasemanagementsystemprovidestheabilityformanydifferentuserstosharedataandprocessresources.Butastherecanbemanydifferentusers,therearemanydifferentdatabaseneeds.Thequestionnowis:Howcanasingle,unifieddatabasemeetthedifferingrequirementofsomanyusers?

ADBMSminimizestheseproblemsbyprovidingtwoviewsofthedatabasedata:anexternalview(orUserview),logicalview(orconceptualview)andphysical(orinternal)view.Theuser’sview,ofadatabaseprogramrepresentsdatainaformatthatismeaningfultoauserandtothesoftwareprogramsthatprocessthosedata.Thatis,thelogicalviewtellstheuser,inuserterms,whatisinthedatabase.Thephysicalviewdealswiththeactual,physicalarrangementandlocationofdatainthedirectaccessstoragedevices(DASDs).Databasespecialistsusethephysicalviewtomakeefficientuseofstorageandprocessingresources.Withthelogicalviewuserscanseedatadifferentlyfromhowtheyarestored,andtheydonotwanttoknowallthetechnicaldetailsofphysicalstorage.Afterall,abusinessuserisprimarilyinterestedinusingtheinformation,notinhowitisstored.

OnestrengthofaDBMSisthatwhilethereistypicallyonlyoneconceptual(orlogical)andphysical(orInternal)viewofthedata,therecanbeanendlessnumberofdifferentExternalviews.Thisfeatureallowsuserstoseedatabaseinformationinamorebusiness-relatedwayratherthanfromatechnical,processingviewpoint.Thusthelogicalviewreferstothewayuserviewsdata,andthephysicalviewtothewaythedataarephysicallystoredandprocessed...

5.2DBMSfeaturesandcapabilities

Alternatively,andespeciallyinconnectionwiththerelationalmodelofdatabasemanagement,therelationbetweenattributesdrawnfromaspecifiedsetofdomainscanbeseenasbeingprimary.Forinstance,thedatabasemightindicatethatacarthatwasoriginally"red"mightfadeto"pink"intime,provideditwasofsomeparticular"make"withaninferiorpaintjob.Suchhigherarityrelationshipsprovideinformationonalloftheunderlyingdomainsatthesametime,withnoneofthembeingprivilegedabovetheothers.

5.3DBMSsimpledefinition

Databasemanagementsystemisthesysteminwhichrelateddataisstoredinan"efficient"and"compact"manner.EfficientmeansthatthedatawhichisstoredintheDBMSisaccessedinveryquicktimeandcompactmeansthatthedatawhichisstoredinDBMScoversverylessspaceincomputer'smemory.Inabovedefinitionthephrase"relateddata"isusedwhichmeansthatthedatawhichisstoredinDBMSisaboutsomeparticulartopic.

Throughoutrecenthistoryspecializeddatabaseshaveexistedforscientific,geospatial,imaging,documentstorageandlikeuses.FunctionalitydrawnfromsuchapplicationshaslatelybegunappearinginmainstreamDBMSsaswell.However,themainfocusthere,atleastwhenaimedatthecommercialdataprocessingmarket,isstillondescriptiveattributesonrepetitiverecordstructures.

Thus,theDBMSsoftodayrolltogetherfrequentlyneededservicesorfeaturesofattributemanagement.ByexternalizingsuchfunctionalitytotheDBMS,applicationseffectivelysharecodewitheachotherandarerelievedofmuchinternalcomplexity.Featurescommonlyofferedbydatabasemanagementsystemsinclude:

5.3.1Queryability

Queryingistheprocessofrequestingattributeinformationfromvariousperspectivesandcombinationsoffactors.Example:"Howmany2-doorcarsinTexasaregreen?"Adatabasequerylanguageandreportwriterallowuserstointeractivelyinterrogatethedatabase,analyzeitsdataandupdateitaccordingtotheusersprivilegesondata.

5.3.2Backupandreplication

Copiesofattributesneedtobemaderegularlyincaseprimarydisksorotherequipmentfails.Aperiodiccopyofattributesmayalsobecreatedforadistantorganizationthatcannotreadilyaccesstheoriginal.DBMSusuallyprovideutilitiestofacilitatetheprocessofextractinganddisseminatingattributesets.Whendataisreplicatedbetweendatabaseservers,sothattheinformationremainsconsistentthroughoutthedatabasesystemanduserscannottellorevenknowwhichserverintheDBMStheyareusing,thesystemissaidtoexhibitreplicationtransparency.

5.3.2Ruleenforcement

Oftenonewantstoapplyrulestoattributessothattheattributesarecleanandreliable.Forexample,wemayhavearulethatsayseachcarcanhaveonlyoneengineassociatedwithit(identifiedbyEngineNumber).Ifsomebodytriestoassociateasecondenginewithagivencar,wewanttheDBMStodenysucharequestanddisplayanerrormessage.However,withchangesinthemodelspecificationsuchas,inthisexample,hybridgas-electriccars,rulesmayneedtochange.Ideallysuchrulesshouldbeabletobeaddedandremovedasneededwithoutsignificantdatalayoutredesign.

5.3.4Security

Oftenitisdesirabletolimitwhocanseeorchangewhichattributesorgroupsofattributes.Thismaybemanageddirectlybyindividual,orbytheassignmentofindividualsandprivilegestogroups,or(inthemostelaboratemodels)throughtheassignmentofindividualsandgroupstoroleswhicharethengrantedentitlements.

5.3.5Computation

Therearecommoncomputationsrequestedonattributessuchascounting,summing,averaging,sorting,grouping,cross-referencing,etc.Ratherthanhaveeachcomputerapplicationimplementthesefromscratch,theycanrelyontheDBMStosupplysuchcalculations.

5.3.6Changeandaccesslogging

Oftenonewantstoknowwhoaccessedwhatattributes,whatwaschanged,andwhenitwaschanged.Loggingservicesallowthisbykeepingarecordofaccessoccurrencesandchanges.

5.3.7Automatedoptimization

Iftherearefrequentlyoccurringusagepatternsorrequests,someDBMScanadjustthemselvestoimprovethespeedofthoseinteractions.InsomecasestheDBMSwillmerelyprovidetoolstomonitorperformance,allowingahumanexperttomakethenecessaryadjustmentsafterreviewingthestatisticscollected

5.4Meta-datarepository

Metadataisdatadescribingdata.Forexample,alistingthatdescribeswhatattributesareallowedtobeindatasetsiscalled"meta-information".Themeta-dataisalsoknownasdataaboutdata.

5.5Currenttrends

In1998,databasemanagementwasinneedofnewstyledatabasestosolvecurrentd