15
Copyright © 2010 Calpont Corporation. All Rights Reserved. Why Choose a Column Database for Business Intelligence? 1 Why Choose a Column Database for Business Intelligence? Calpont Corporation 3011 Internet Blvd, Suite 100 Frisco, TX 75034 www.calpont.com [email protected]

Why Choose a Column Database for Business Intelligence

Embed Size (px)

DESCRIPTION

Column database

Citation preview

  • Copyright2010CalpontCorporation.AllRightsReserved.

    WhyChooseaColumnDatabaseforBusinessIntelligence?

    1

    WhyChooseaColumnDatabaseforBusinessIntelligence?

    CalpontCorporation3011InternetBlvd,Suite100

    Frisco,[email protected]

  • Copyright2010CalpontCorporation.AllRightsReserved.

    WhyChooseaColumnDatabaseforBusinessIntelligence?

    2

    WhyChooseaColumnDatabaseforBusinessIntelligence?November2010Copyright2010CalpontCorporation.AllRightsReserved.CalpontandInfiniDBaretrademarksofCalpontCorporation.Otherproductnamesaretrademarksoftheirrespectiveowners.

  • Copyright2010CalpontCorporation.AllRightsReserved.

    WhyChooseaColumnDatabaseforBusinessIntelligence?

    3

    TableofContents

    ExecutiveSummary ...................................................................................................................................................4HowDoColumnOrientedDatabasesWork?............................................................................................................5DoColumnOrientedDatabasesReallyMakeaDifference?.....................................................................................6

    ABenchmarkComparisonbetweenInfiniDBandaLeadingRowBasedDatabase ..............................................8AdditionalConsiderations ...................................................................................................................................10

    TheLearningCurvewithColumnOrientedDatabases ...........................................................................................11TheExpenseofImplementingColumnOrientedDatabases11

    GeneralWaysinWhichColumnDatabasesHelpSaveMoney ...........................................................................10TheInfiniDBAdvantage...........................................................................................................................................12

    ExploitsTodaysModernHardware ....................................................................................................................13MPPCapable.......................................................................................................................................................13SelfManagingDatabase ....................................................................................................................................14HighSpeedDataLoading....................................................................................................................................14HighDataConcurrencyandTransactionalDataProtection ...............................................................................15MySQLInterface..................................................................................................................................................15ExtensiveApplicationDevelopmentandThirdPartySoftwareSupport .............................................................15

    Conclusion ...............................................................................................................................................................15

  • Copyright2010CalpontCorporation.AllRightsReserved.

    WhyChooseaColumnDatabaseforBusinessIntelligence?

    4

    ExecutiveSummarySometechnologiescomeontheinformationtechnologylandscapeandstay,providinglonglastingbenefits,whereasothersaremoreofashorttermfadandultimatelyendupdisappearingbecausethevaluetheysuppliedwastoonicheorientedand/ortheywerequicklysupplantedbyanothertechnologythatisbetter.Recently,articles,blogs,analystreports,andothermediaoutletshavebeennotingthemeteoricriseandusageofcolumnorienteddatabasesintheareasofdatawarehousing,analytics,andotherbusinessintelligencesituations.Arecolumnorienteddatabasesatechnologythatisdestinedtostayandprovidelongtermbenefitsorwillitberelegatedtotheforgottenpileofothersoftwarethatcameonthescenequicklyandthendisappeared?Onethingseemsclear:thekeyareathatcolumnorienteddatabasesservebusinessintelligence(BI)seemsheretostay.AJuly2009studybytheAberdeengroupfoundthatbusinessintelligenceisexpectedtohavethehighestimpactonorganizationsoverthenexttwotofiveyears.Insurveyingmorethan1,600executivesin36countries,AberdeenfoundthatoneinfouridentifiedBIandanalyticsasthesoftwaretechnologywiththemostpronouncedeffectin2009.1AgreeingwithAberdeenisIBMwhofoundinarecentstudythat83%ofexecutivesnowrankbusinessintelligenceandanalyticsastheirtoppriority,2aswellasGartnerGroupGartnerGroupwhofoundthatin2009,businessintelligence(BI)rankedasthetoptechnologypriorityforthechiefinformationofficers.3ButhowexactlycancolumnorienteddatabaseshelpITdirectorsandBIarchitectswhoarechargedwithimplementingmoderndatawarehouses,businessintelligencedatabases,andotherreadintensiveapplications?Thispaperexaminesfourkeyquestionsthatareconsistentlyaskedofcolumnorienteddatabases:

    1. Howdocolumnorienteddatabaseswork?2. Docolumnorienteddatabasesreallymakeadifference?3. Whatlearningcurve(application/databasedevelopment,etc.)isinvolvedwithcolumn

    orienteddatabases?4. Whatstheexpenseofimplementingcolumnorienteddatabases?

    Thispaperalsohighlights,anddiscussesthefeaturesandbenefitsofthecolumnorienteddatabasefromCalpont,InfiniDB,anddescribeshowitcanaddressthedatabaserequirementsofanyBIapplication.

    1AntoneGonsalves,BIisaTopPriorityonExecutivesWishLists,InformationWeek,July2009.2DougHenschen,NextStepsforAnalyticsintheBigDataEra,IntelligentEnterprise,November2009.3http://www.webwire.com/ViewPressRel.asp?aId=87637.

  • Copyright2010CalpontCorporation.AllRightsReserved.

    WhyChooseaColumnDatabaseforBusinessIntelligence?

    5

    HowDoColumnOrientedDatabasesWork?Allthelegacyrelationaldatabasescurrentlybeingofferedtodaywereandareprimarilydesignedtohandleonlinetransactionalprocessing(OLTP)workloads.Atransaction(e.g.anonlineorderforabookthroughAmazonoranotherWebbasedbookdealer)typicallymapstooneormorerowsinarelationaldatabase,andalltraditionalRDBMSdesignsarebasedonaperrowparadigm.Fortransactionalbasedsystems,thisarchitectureiswellsuitedtohandletheinputofincomingdata.However,forapplicationsthatareveryreadintensiveandselectiveintheinformationbeingrequested,theOLTPdatabasedesignisntamodelthattypicallyholdsupwell.Whereastransactionsarerowbased,mostdatabasequeriesarecolumnbased.Insertinganddeletingtransactionaldataarewellservedbyarowbasedsystem,butselectivequeriesthatareonlyinterestedinafewcolumnsofatablearehandledmuchbetterbyacolumnorientedarchitecture.Onaverage,arowbasedsystemdoes510xthephysicalI/Othatacolumnbaseddatabasedoestoretrievethesameinformation.TakingintoaccountthatphysicalI/Oistypicallytheslowestpartofaquery,andthatananalyticalquerytypicallytouchessignificantlymorerowsofdatathatatypicaltransactionaldatabaseoperation,theperformancegapbetweenroworientedarchitecturesandcolumnorientedarchitectureoftentimeswidensasthedatabasegrows.Togetaroundtheirselectivequeryinefficiencies,rowbasedRDBMSsutilizeindexing,horizontalpartitioning,materializedviews,summarytables,andparallelprocessing,allofwhichcanprovidebenefitsforintensivequeries,buteachcomeswiththeirownsetofdrawbacksaswell.Forexample,whileindexingcancertainlyhelpqueriescompletefasterinsomecases,theyalsorequiremorestorage,impedeinsert/update/deleteandbulkloadoperations(becausetheindexesmustbemaintainedaswellastheunderlyingtable),andcanactuallydegradeperformancewhentheybecomeheavilyfragmented.Moreover,inbusinessintelligence/analyticenvironments,theadhocnatureofsuchscenariosmakesitnearlyimpossibletopredictwhichcolumnswillneedindexing,sotablesendupeitherbeingoverindexed(whichcausesloadandmaintenanceissues)ornotproperlyindexedandsomanyqueriesenduprunningmuchslowerthandesired.ManyITprofessionalshavealreadyrecognizedthesuperiorityofcolumndatabasesinreadintensiveenvironmentsandhavestartedusingthem.Attheendof2009,aTDWIresearchstudyfoundthatfiftyonepercentofitsrespondentsalreadyuseacolumnorienteddatabasenow,withthatnumberprojectedtoincreaseto75%inthreeyears.4Whysuchmomentumforcolumndatabases?ITprofessionalsarefindingoutthat,forbusinessintelligencedatabaseactivities,theysimplyworkasadvertised.ThosenotfamiliarwithacolumnorienteddatabasemightwonderexactlywhattheyareandwhatactualbenefitstheydeliveroveralegacyRDBMS.Itsimportanttonotethat,onthesurface,acolumnorienteddatabaseappearsexactlylikeatraditionalrelationaldatabase:thelogicalconceptsoftablesandrowsarethesame,SQLcommandsareusedtointeractwiththesystem,andmostotherRDBMSparadigms(e.g.security,backup/recovery,etc.)remainunchanged.But,acolumnorienteddatabasespecificallydesignedforanalyticsovercomesthequerylimitationsthatexistintraditionalRDBMSsystemsbystoring,managing,andqueryingdatabasedoncolumnsratherthan

    4PhilipRussom,NextGenerationDataWarehousePlatforms,TDWI,2009.

  • Copyright2010CalpontCorporation.AllRightsReserved.

    WhyChooseaColumnDatabaseforBusinessIntelligence?

    6

    rows.Becauseonlythenecessarycolumnsinaqueryareaccessedratherthanentirerows,I/Oactivitiesaswellasoverallqueryresponsetimescanbereduced.Inotherwords,ifyoudonthavetoreadanentirerowtogetthedatayouneed,whydoit?Theendresultforcolumndatabasesistheabilitytointerrogateandreturnqueryresultsagainsteithermoderateamountsofinformation(tensorhundredsofGBs)orlargeamountsofdata(1nterabytes)inmuchlesstimethatstandardRDBMSsystemscan.

    DoColumnOrientedDatabasesReallyMakeaDifference?ManyBIapplicationscurrentlymakeuseoflegacyRDBMSservers,butthattrendisrapidlychanging.TDWIfoundanegativetrend(specifically52%)whenitdidamajorstudyandaskedthequestioniftraditionalrowbaseddatabasesweregoingtobeusedforBIimplementations.5ThisdevelopmentisnotsurprisingbecauseexperiencedBIprofessionalshaveknownforalongtimethatlegacyRDBMSsarenotdesignedforbusinessintelligenceworkloads.Theyperformverywellfortransactionprocessingsystems,butwhenitcomestodatabaseanalytics,legacydatabasevendorproductsdonotdelivereitherthespeedorthepricetagthatthoserollingoutmodernBIapplicationsneed.ThesameTDWIstudyfoundthatthemostimportantcomponentinabusinessintelligenceimplementationwasthedatabaseserveritself.6

    5Russom,ibid.6Russom,ibid.

  • Copyright2010CalpontCorporation.AllRightsReserved.

    WhyChooseaColumnDatabaseforBusinessIntelligence?

    7

    Nearlythreequartersofthestudysrespondentsindicatedthatitisthedatabasesoftware,whichisthemostcriticalpieceofaBIapplicationssuccess.Further,TDWIfoundthatnearlyhalfofthoseitpolledarereadytoreplacetheirdatabaseusedforbusinessintelligenceapplicationswithanother,moremodernalternative.Whenaskedwhatthetechnicalreasonswereforthereplacement,thenumberoneanswerwastheinabilityforthelegacyRDBMStoservicequeriesinthetimeneeded:

    CitingarecentstudydonebyVentana,datawarehouseauthorDavidStodderbringstolightthatnearly60%ofdatawarehouseimplementationsareexperiencingnightmarishperformanceproblemswhentheyhavetoscaleandruncomplexqueries,andnearly50%saytheyhavethesameproblemwhenitcomestoloadingmoredata.TheendresultsareBIqueriesbeingcanceledbecausetheysimplyruntoolong.Because

  • Copyright2010CalpontCorporation.AllRightsReserved.

    WhyChooseaColumnDatabaseforBusinessIntelligence?

    8

    ofthis,theVentanastudyfoundtheseorganizationsarenowseriouslyevaluatingcolumndatabasestosolvetheirproblems.7Notingthatdatavolumesaregrowingatarampantratewithinmostenterprises(withestimatesrangingfrom20percentperyearontheconservativesideto50percentperyearamongthelargestorganizations),ForresterResearchsaysthattheprognosisis"withinafewyears,traditionalroworientedrelationaldatabasesarereallygoingtobepushingtheirlimits.TherearealotofDBAswhodontwanttohearitbecausethey'resoentrenchedinOracleandIBMDB2andMicrosoftSQLServer,butwethinkaneweraiscoming."8DocolumndatabasesreallyhavethecapabilitytodeliverwhatITexecutivesneed?CommentingonthetechnologybenefitofcolumnorienteddatabasesisGartnerGroupsdatabaseexpertDonaldFeinbergwhosays,"Ifyou'rebringingbackallthecolumns,acolumnstoredatabaseisn'tgoingtoperformanybetterthanarowstoreDBMS,butanalyticapplicationsaretypicallylookingatallrowsandonlyafewcolumns.WhenyouputthattypeofapplicationonacolumnstoreDBMS,itoutperformsanythingthatdoesn'ttakeacolumnstoreapproach."9

    ABenchmarkComparisonbetweenInfiniDBandaLeadingRowBasedDatabase AsanexampleofhowacolumnorienteddatabasecanoutperformalegacyRDBMS,Calpontrecentlycommissionedawellknowndatawarehouseindustryexperttobenchmarktheleadingrowbaseddatabase(ofwhichtheexperthasmanyyearsofexperienceintuningforfastperformance)againstInfiniDB,whichhasasoneofitscorefeatures,acolumnorienteddesign.TheStarSchemastyledbenchmarkwasconductedontwodifferentmachinestogaugeperformanceonbothmidandlargesizedservers.Themidsizedserverwasan8CPU,8GBRAM,14SATA7200RAID0nocacheconfiguration,andthelargeserverwasa16CPU,16GBRAM,14SAS15KRPMRAID0with512MBcachemachine.Bothwererunning64bitCentOS5.4.Therawdatabasesizewas2TB.Ascanbeseenonthegraphsbelow,variousconfigurationswereusedfortheleadingrowbaseddatabase,howevernomattertheconfiguration,thecolumnorientedInfiniDBdatabaseconsistentlyanddramaticallybeatthelegacydatabaseinstoragefootprint,loadtime,andqueryspeed:

    7DavidStodder,NineBIMegatrendsfor2009,IntelligentEnterprise,January,2009.8DougHenschen,ColumnStoreDatabasesandDWAppliances:HowtoMaketheRightChoice,InformationWeekResearchandReports,March2008.9Henschen,ibid.

  • Copyright2010CalpontCorporation.AllRightsReserved.

    WhyChooseaColumnDatabaseforBusinessIntelligence?

    9

  • Copyright2010CalpontCorporation.AllRightsReserved.

    WhyChooseaColumnDatabaseforBusinessIntelligence?

    10

    Insummary,theInfiniDBserversavesonstoragecosts,suppliesfasteraccesstonew/incomingdata,andrunsquerymuchfasterthanitsrowbasedcompetitor.Inadditiontoproducingoverallfasterqueryspeeds,InfiniDBalsosuppliedmuchbetterquerypredictabilityintermsofquerytime.Whereastheleadingrowbaseddatabaseproducedwildlyvaryingminimumandmaximumquerytimesoverthevariousruns,InfiniDBhadafarmoretightlygroupofrunswhenitcametopredictableresponsetimes.ThistranslatesintomuchbetterdependabilityfromabusinessstandpointinensuringBIreportsandqueriesmeetwhateverservicelevelagreementsareimposedfrombusinessusers.

    AdditionalConsiderationsInadditiontobetterperformance,thecolumnorientationaspectofcolumndatabasessuppliesanumberofusefulbenefitstothosewishingtodeployfastbusinessintelligencedatabases.First,thereisnoneedforindexingaswithtraditionalrowbaseddatabases.Theeliminationofindexingmeans:(1)lessoverallstorageisconsumedincolumndatabasesbecauseindexesinlegacyRDBMSsoftenballoonthestoragecostofadatabasetodoubleormoretheinitialdatasize;(2)dataloadspeedisincreasedbecausenoindexesneedtobemaintained;(3)adhocDMLworkspeedisincreasedbecausenoindexupdatesareperformed;(4)noindexingdesignortuningworkisimposedonthedatabaseITstaff.Second,thereisfarlessdesignworkforcedondatabasearchitectswhencolumndatabasesareused.Theneedforcomplicatedpartitioningschemes,materializedvieworsummarytabledesigns,andothersuch

  • Copyright2010CalpontCorporation.AllRightsReserved.

    WhyChooseaColumnDatabaseforBusinessIntelligence?

    11

    workiscompletelyremovedbecausecolumndatabasesneednoneofthesecomponentstoachievesuperiorqueryperformance.

    TheLearningCurvewithColumnOrientedDatabasesBothITmanagementanddatabasedevelopers/administratorswillbepleasedtofindthatthelearningcurveassociatedwithmovingfromlegacy,rowbasedRDBMSstoacolumndatabaseisverysmallifnotcompletelynonexistent.Unlikeotherdatabasesthatcameonthesceneinprioryearswhichrequiredeitherdifferentprogrammingparadigms(e.g.objectorienteddatabases)orlearningnewdesignmethodologiesanddatabaseaccesslanguages(e.g.OLAPdatabases),columndatabaseslookandhandlejustlikestandardrelationaldatabases.TheyusethesameANSIstandardSQLlanguage,securitymethods,andrequirenodevelopmentparadigmchanges.Infact,columndatabasesactuallylessentheburdenonboththedevelopmentandadministrationstaffbecausetheydoawaywiththeneedforindexingexercises,datapartitioningschemes,supplementaryobjectdesigns(e.g.materializedviews),andothersimilartasks.Theeaseofusefactor,therefore,isgreaterwithcolumndatabasesthanitiswithtraditionalRDBMSs.Moreover,theydonotrequiresuchspecializedinhouseexpertisetobuildhighlyperformantsystems.

    TheExpenseofImplementingColumnOrientedDatabasesAnInformationWeeksurveyfoundthatnearly40%ofITexecutivescitedexpensivesoftwarelicensingcostsasthenumberonereasontheyareunabletorolloutalltheBIinitiativestheyrequire,andastudyperformedbyTDWIfoundthatalmost60%ofITmanagershavecitedrecenteconomicconditionsasthereasonforconductingareevaluationoftheirsoftwarevendorsinanefforttoreducecostsandimplementtheBIsystemstheirbusinessneeds.10Inaddition,therecentglobalrecessionhascausedwhatmanybelievetobeapermanentchangeinmindsetawayfromhighpricedproprietaryhardwareandsoftwaretocommoditytechnologiesthatareoftentimesfoundtobenearlytechnicallyequal,andsometimesactuallysuperiorintermsofoverallefficiencyandperformance.TDWIfoundthatnearly60%ofitsrespondentsarenowconsideringlowercostoptionsfortheirbusinessintelligencedatabase.Reasonsforthisincludedreducedbudgets,frozenhiring,projectsbeingputonhold,softwareacquisitiondenials,andstaffreductions.11

    GeneralWaysinwhichColumnDatabasesHelpSaveMoneyColumnorienteddatabasesandthetechnologythatisbeneficiallyutilizedbythemmaycertainlysoundenticing,butthequestionneedstobeasked:howcancolumndatabaseshelpcontainITcosts?Hereareafewgeneralwayscolumndatabaseshelpdefrayexpenses:

    Mostallcolumndatabasesexploittheadvancedincommodityhardware(e.g.64bit,multiCPU/core)anddonotrequireexpensivemachinesbeusedtoachieveexceptionalperformance.

    10PhilipRussom,NextGenerationDataWarehousePlatforms,TDWI,2009.11Russom,ibid.

  • Copyright2010CalpontCorporation.AllRightsReserved.

    WhyChooseaColumnDatabaseforBusinessIntelligence?

    12

    ForcolumndatabasesthatsupportMPP,thepriceperformancemetric(takingcommodityhardwareintoaccount)canbecomequiteimpressive

    Themajorityofcolumndatabasesoffersomeformofdatacompressioneitherphysicalorlogicalthathelpssavestoragecosts

    Inadditiontogeneraldatacompression,manycolumndatabasesdonotrequiretheuseofindexes,materializedviews,orothersupplementarystoragestructuresthatareusedtohelp

    theperformanceoftraditionalRDBMSs.Becausethesestructuresarenotused,additionalstoragecostsavingsaretheendresult

    Becausecolumndatabasesdonotrequireindexesorcomplicateddatabasedesignschemes(e.g.datapartitioning,etc.)anddomuchoftheperformancedesignworkautomaticallyforthe

    enduser,oftentimestheneedforhighlyexperienceddatabasestaffisnegated.ThismeansthatinexperiencedITstaffcancreatebusinessintelligencedatabasesthatperformextremelywellwithouthavingtopossesssophisticatedknowledgeofdatawarehousedesignsandperformancetuningtechniques

    Theincreasedspeedatwhichcolumndatabasesoperatetypicallymeansthatmorebusinessintelligencereports/requestscanbedeliveredanddecisionmakerscanmorequicklyperformtheirjob,whichlessensstaffidletime

    Commentingoncolumndatabasesandcostsavings,GartnersFeinbergsays,InthetwospecificareaswherecolumnstoreDBMSexcel(analyticsandarchivingsolutions),therecanbealargecostsavingsand,withanalyticapplications,alargeperformancegain.

    TheInfiniDBAdvantageInfiniDBisacolumnorienteddatabasethatisdesignedtoservicetheneedsofbusinessintelligenceapplicationsaswellasanyotherscenariowheremediumtoverylargeamountsofdataneedtobequeriedandfastresponsetimesaredesired.TheapplicationsandsystemsthatwillbenefitfromInfiniDBaremany,withtherebeingthreecommonusecaseswhereInfiniDBcanbeexpectedtoexcel:

    1. Datawarehouses,datamarts,andotherbusinessintelligence(BI)datastores2. Generalpurposereportingdatabases3. Readintensivesegmentsofanoverallapplicationwhereread/search/lookupportionsofthe

    applicationareservedbyananalytic/querydatabaseandthetransactionalsegments(e.g.orderentry,etc.)aremanagedbyatraditionalrelationaldatabase

    Inadditiontobeingacolumnorienteddatabase,InfiniDBoffersanumberofotherbenefitsthattodaysbusinessintelligenceuserswantintheirunderlyingdatabase.Someofthecorefeaturesandbenefitsarelistedbelow.

  • Copyright2010CalpontCorporation.AllRightsReserved.

    WhyChooseaColumnDatabaseforBusinessIntelligence?

    13

    ExploitsTodaysModernHardwareInfiniDBiscapableofexploitingmodern64bitcommodityhardwareandutilizingthemanyCPUsandcoresthatcomestandardontodaysservers.Becauseofitsmultithreadeddesign,InfiniDBcanscaleuponSMPmachinesandparallelizeitsoperationsacrossthemachinesresourcestoproduceveryfastqueryresponsetimes.NotethatInfiniDBcanbeusedon32bitmachinesaswellifdesired.

    MPPCapableInadditiontobeingabletoscaleup,InfiniDBcanalsoscaleoutinmassiveparallelprocessing(MPP)fashionacrossmultiplecommodityhardwaremachinestoachievelinearincreasesinoverallperformance.AddinginexpensivehardwaretoanInfiniDBconfigurationallowsthedatabasetoincreaseprocessingpowersoresponsetimescanmanytimesbereducedinhalfjustwiththeadditionofanewnode.Thismakesiteasytoadjustforgrowingdatavolumes,increaseduseractivity,orthedesiretomeetbetteroverallperformancegoals.Moreover,thiscanbeaccomplishedinadynamicfashionwheretheInfiniDBsystemdoesntgodownorneedtobetakenofflinewhennewnodesareadded.InfiniDBsmodulararchitectureallowsateamchargedwithdevelopingafastbusinessintelligencedatabasetoscalebothI/Oanduserconcurrency.NewnodescanbeaddedbothattheI/O(PerformanceModules)anduserconnectionlayer(UserModules)toensurenobottleneckinprocessingisencountered:

    InfiniDBsMPPcapabilitiesofferanotherbenefitthatofadistributedsharednothingdatacache.InamultiplenodeInfiniDBconfiguration,dataisdistributedamongthevariousnodesandtheirdatacaches.Nonodesharesdatawiththeother,howeverallareaccessedintheInfiniDBMPParchitecturewhendataisreadtosatisfyqueries.Inessencethen,InfiniDBcreatesonelargelogicaldatacachethatisaccessedinadistributedfashioninparallelbyallparticipatingnodes.ThisallowsInfiniDBtoliterallycachelargedatabaseswhenenoughnodesarepresentwithgenerousamountsofmemory.

  • Copyright2010CalpontCorporation.AllRightsReserved.

    WhyChooseaColumnDatabaseforBusinessIntelligence?

    14

    Ofcourse,InfiniDBcanrunonastandalonemachinewithlocaldiskstoragequitewellandofferscaleupperformanceutilizingthatserversCPUsandotherresources.ButInfiniDBisnotlimitedtobeingdeployedonsinglemachines,andiswellsuitedforMPPprocessing.SelfManagingDatabase

    GettingexceptionaldataloadandqueryperformancefromInfiniDBdoesnotrequireseasoneddatabaseexpertstodesignandtunethesystem.InfiniDBtakesthedatabasedesignandadministrationheadachesawayfrombusinessintelligenceprofessionalsthroughitsautomanagementarchitectureframework.First,beingcolumnoriented,InfiniDBrequiresnoindexing,whichmeansdatabaseadministratorsanddevelopersdonthavetoworryaboutcomplexindexingschemesandthestorageoverheadsuchstructuresimpose.Beingcolumnoriented,InfiniDBobviatestheneedfortraditionalindexingbecause,inessence,thecolumnistheindex.Second,InfiniDBremovestheneedformanualdatapartitioningandstorageplacement.InfiniDBusesaformoflogicalhorizontalrangepartitioningthatdoesnotrequirespecialstorageplacementorschemadesign.Usingbothvertical(columnorientation)andlogicalhorizontalrangepartitioningallowsInfiniDBtoreduceI/Oinbothdirections(columnandrow).BothverticalandhorizontalpartitioningareautomaticallyhandledbytheInfiniDBdatabaseandrequirenouserintervention.Finally,whendeployedinanMPPfashion,InfiniDBoffersautomaticI/OandconcurrencyscalingplusautomaticfailoverattheI/Olayer.Inaddition,patchmanagementofallparticipatingInfiniDBnodesiseasedthroughanautomatedsoftwareprovisioningcapability.WhenpatchesorupgradesneedtobeappliedtomorethanoneInfiniDBserver,anautomaticprocesstakesthesoftwarefromthefirstnodewhereanupgradeisappliedandautomaticallyupgradesallotherparticipatingnodes.HighSpeedDataLoadingGettingdataloadedintobusinessintelligencedatabasesasquicklyaspossibleisahighpriorityinmostcases.InfiniDBoffersahighspeed,bulkloadutilitythatloadsdataveryquicklyintoInfiniDBtables.Becausetherearenoindexes,loadtimesarereducedevenfurther.InlegacyRDBMSs,astandardpracticeistodropindexesontablessodataloadscanrunfasterandnewdatacanbeaddedtothesystemmorequickly.However,therearetwoproblemswiththisapproach:(1)theindexeswillhavetoberecreated,whichusesupCPUandmemoryresourcesonthesystem;(2)anyqueriesthatneedtoberunduringthistimewillrunmuchslowerbecausenoindexeswillbeavailable,whichequatestofulltablescansonlargetablesnotsomethinglegacydatabasesareknownforhandlingwell.ButtheseissuesarenonexistentinInfiniDBsinceindexingisunnecessary.Inadditiontoahighspeedbulkloader,InfiniDBsupportsanumberofotherMySQLbasedloadutilities,ETLtools,andstandardDML(insert,update,delete)commands.

  • Copyright2010CalpontCorporation.AllRightsReserved.

    WhyChooseaColumnDatabaseforBusinessIntelligence?

    15

    HighDataConcurrencyandTransactionalDataProtection

    InfiniDBemploysanMVCC(multiversionconcurrencycontrol)designsoreadersneverblockwritersordataloadsandviceversa.Aquerywillalwaysseethedataasitexistedatthetimethequerywasissued.Inaddition,ACIDcomplianttransactionalsupportisprovidedintheInfiniDBdatabase.Transactionscaneasilybecommittedorrolledback,anddeadlockdetectionsupportisalsoprovidedtohandleconflictresolution.Finally,InfiniDBprovidesfullcrashrecoverycapabilities.Intheeventofasystemcrash,InfiniDBautomaticallymaintainsdataintegrityanduponsystemrestart,InfiniDBsupportsrollforwardandrollbackoperationstoreturnthedatabasetoaconsistentstate.

    MySQLInterface

    InfiniDButilizesMySQLforitsbasicfrontenduserinterface.ThisallowsanyonefamiliarwithMySQLtobecomeimmediatelyproductivewithInfiniDB.ForthosenotacquaintedwithMySQL,thelearningcurveisminimalasMySQLsupportsalmostallANSIstandardSQLoperations.Moreover,therearemanyfreelysuppliedGUItoolsfromMySQLaswellasothervendorsthatmaybeusedtodevelopagainstandadministeranInfiniDBdatabase.

    ExtensiveApplicationDevelopmentandThirdPartySoftwareSupportInfiniDBoffersmanydifferentdriversandAPIs(ODBC,JDBC,PHP,etc.)forconnectingtothedatabase,aswellasplentyofSQLandapplicationdevelopmentaidsthatarebuiltintotheserver.Moreover,anythirdpartysoftwarebusinessintelligencetool(oranysoftwareproductforthatmatter)thatsupportsMySQLshouldworkwellwithInfiniDB.

    ConclusionIntheend,ITorganizationsshouldconsideracolumndatabaseoveralegacyRDBMSbecauseitcanmakeabigimpactinhowbusinessintelligencesystemsperformandthedownstreamimpactthiscanhaveinthetimelinessofdecisionmaking.

    CalpontInfiniDBTheHighPerformance,ScalableDataPlatformthatenablesDeepAnalytics

    FormoreinformationortopurchaseCalpontInfiniDB,pleasecontact2146189500,orvisitwww.calpont.com.