You are on page 1of 15

WhyChooseaColumnDatabaseforBusinessIntelligence?

WhyChooseaColumnDatabasefor
BusinessIntelligence?

CalpontCorporation
3011InternetBlvd,Suite100
Frisco,TX75034
www.calpont.com
info@calpont.com

Copyright2010CalpontCorporation.AllRightsReserved.

WhyChooseaColumnDatabaseforBusinessIntelligence?

WhyChooseaColumnDatabaseforBusinessIntelligence?
November2010

Copyright2010CalpontCorporation.AllRightsReserved.

CalpontandInfiniDBaretrademarksofCalpontCorporation.Otherproductnamesaretrademarksoftheirrespectiveowners.

Copyright2010CalpontCorporation.AllRightsReserved.

WhyChooseaColumnDatabaseforBusinessIntelligence?

TableofContents

ExecutiveSummary ...................................................................................................................................................4
HowDoColumnOrientedDatabasesWork?............................................................................................................5
DoColumnOrientedDatabasesReallyMakeaDifference?.....................................................................................6
ABenchmarkComparisonbetweenInfiniDBandaLeadingRowBasedDatabase ..............................................8
AdditionalConsiderations ...................................................................................................................................10
TheLearningCurvewithColumnOrientedDatabases ...........................................................................................11
TheExpenseofImplementingColumnOrientedDatabases11
GeneralWaysinWhichColumnDatabasesHelpSaveMoney ...........................................................................10
TheInfiniDBAdvantage...........................................................................................................................................12
ExploitsTodaysModernHardware ....................................................................................................................13
MPPCapable.......................................................................................................................................................13
SelfManagingDatabase ....................................................................................................................................14
HighSpeedDataLoading....................................................................................................................................14
HighDataConcurrencyandTransactionalDataProtection ...............................................................................15
MySQLInterface..................................................................................................................................................15
ExtensiveApplicationDevelopmentandThirdPartySoftwareSupport .............................................................15
Conclusion ...............................................................................................................................................................15

Copyright2010CalpontCorporation.AllRightsReserved.

WhyChooseaColumnDatabaseforBusinessIntelligence?

ExecutiveSummary
Sometechnologiescomeontheinformationtechnologylandscapeandstay,providinglonglastingbenefits,
whereasothersaremoreofashorttermfadandultimatelyendupdisappearingbecausethevaluethey
suppliedwastoonicheorientedand/ortheywerequicklysupplantedbyanothertechnologythatisbetter.
Recently,articles,blogs,analystreports,andothermediaoutletshavebeennotingthemeteoricriseand
usageofcolumnorienteddatabasesintheareasofdatawarehousing,analytics,andotherbusiness
intelligencesituations.Arecolumnorienteddatabasesatechnologythatisdestinedtostayandprovide
longtermbenefitsorwillitberelegatedtotheforgottenpileofothersoftwarethatcameonthescene
quicklyandthendisappeared?

Onethingseemsclear:thekeyareathatcolumnorienteddatabasesservebusinessintelligence(BI)
seemsheretostay.AJuly2009studybytheAberdeengroupfoundthatbusinessintelligenceisexpected
tohavethehighestimpactonorganizationsoverthenexttwotofiveyears.Insurveyingmorethan1,600
executivesin36countries,AberdeenfoundthatoneinfouridentifiedBIandanalyticsasthesoftware
technologywiththemostpronouncedeffectin2009.1AgreeingwithAberdeenisIBMwhofoundina
recentstudythat83%ofexecutivesnowrankbusinessintelligenceandanalyticsastheirtoppriority,2as
wellasGartnerGroupGartnerGroupwhofoundthatin2009,businessintelligence(BI)rankedasthetop
technologypriorityforthechiefinformationofficers.3

ButhowexactlycancolumnorienteddatabaseshelpITdirectorsandBIarchitectswhoarechargedwith
implementingmoderndatawarehouses,businessintelligencedatabases,andotherreadintensive
applications?Thispaperexaminesfourkeyquestionsthatareconsistentlyaskedofcolumnoriented
databases:

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

Thispaperalsohighlights,anddiscussesthefeaturesandbenefitsofthecolumnorienteddatabasefrom
Calpont,InfiniDB,anddescribeshowitcanaddressthedatabaserequirementsofanyBIapplication.

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

Copyright2010CalpontCorporation.AllRightsReserved.

WhyChooseaColumnDatabaseforBusinessIntelligence?

HowDoColumnOrientedDatabasesWork?
Allthelegacyrelationaldatabasescurrentlybeingofferedtodaywereandareprimarilydesignedtohandle
onlinetransactionalprocessing(OLTP)workloads.Atransaction(e.g.anonlineorderforabookthrough
AmazonoranotherWebbasedbookdealer)typicallymapstooneormorerowsinarelationaldatabase,
andalltraditionalRDBMSdesignsarebasedonaperrowparadigm.Fortransactionalbasedsystems,this
architectureiswellsuitedtohandletheinputofincomingdata.

However,forapplicationsthatareveryreadintensiveandselectiveintheinformationbeingrequested,the
OLTPdatabasedesignisntamodelthattypicallyholdsupwell.Whereastransactionsarerowbased,most
databasequeriesarecolumnbased.Insertinganddeletingtransactionaldataarewellservedbyarow
basedsystem,butselectivequeriesthatareonlyinterestedinafewcolumnsofatablearehandledmuch
betterbyacolumnorientedarchitecture.Onaverage,arowbasedsystemdoes510xthephysicalI/Othat
acolumnbaseddatabasedoestoretrievethesameinformation.TakingintoaccountthatphysicalI/Ois
typicallytheslowestpartofaquery,andthatananalyticalquerytypicallytouchessignificantlymorerows
ofdatathatatypicaltransactionaldatabaseoperation,theperformancegapbetweenroworiented
architecturesandcolumnorientedarchitectureoftentimeswidensasthedatabasegrows.

Togetaroundtheirselectivequeryinefficiencies,rowbasedRDBMSsutilizeindexing,horizontal
partitioning,materializedviews,summarytables,andparallelprocessing,allofwhichcanprovidebenefits
forintensivequeries,buteachcomeswiththeirownsetofdrawbacksaswell.Forexample,whileindexing
cancertainlyhelpqueriescompletefasterinsomecases,theyalsorequiremorestorage,impede
insert/update/deleteandbulkloadoperations(becausetheindexesmustbemaintainedaswellasthe
underlyingtable),andcanactuallydegradeperformancewhentheybecomeheavilyfragmented.
Moreover,inbusinessintelligence/analyticenvironments,theadhocnatureofsuchscenariosmakesit
nearlyimpossibletopredictwhichcolumnswillneedindexing,sotablesendupeitherbeingoverindexed
(whichcausesloadandmaintenanceissues)ornotproperlyindexedandsomanyqueriesenduprunning
muchslowerthandesired.

ManyITprofessionalshavealreadyrecognizedthesuperiorityofcolumndatabasesinreadintensive
environmentsandhavestartedusingthem.Attheendof2009,aTDWIresearchstudyfoundthatfiftyone
percentofitsrespondentsalreadyuseacolumnorienteddatabasenow,withthatnumberprojectedto
increaseto75%inthreeyears.4Whysuchmomentumforcolumndatabases?ITprofessionalsarefinding
outthat,forbusinessintelligencedatabaseactivities,theysimplyworkasadvertised.

Thosenotfamiliarwithacolumnorienteddatabasemightwonderexactlywhattheyareandwhatactual
benefitstheydeliveroveralegacyRDBMS.Itsimportanttonotethat,onthesurface,acolumnoriented
databaseappearsexactlylikeatraditionalrelationaldatabase:thelogicalconceptsoftablesandrowsare
thesame,SQLcommandsareusedtointeractwiththesystem,andmostotherRDBMSparadigms(e.g.
security,backup/recovery,etc.)remainunchanged.

But,acolumnorienteddatabasespecificallydesignedforanalyticsovercomesthequerylimitationsthat
existintraditionalRDBMSsystemsbystoring,managing,andqueryingdatabasedoncolumnsratherthan
4

PhilipRussom,NextGenerationDataWarehousePlatforms,TDWI,2009.
Copyright2010CalpontCorporation.AllRightsReserved.

WhyChooseaColumnDatabaseforBusinessIntelligence?

rows.Becauseonlythenecessarycolumnsinaqueryareaccessedratherthanentirerows,I/Oactivitiesas
wellasoverallqueryresponsetimescanbereduced.Inotherwords,ifyoudonthavetoreadanentire
rowtogetthedatayouneed,whydoit?

Theendresultforcolumndatabasesistheabilitytointerrogateandreturnqueryresultsagainsteither
moderateamountsofinformation(tensorhundredsofGBs)orlargeamountsofdata(1nterabytes)in
muchlesstimethatstandardRDBMSsystemscan.

DoColumnOrientedDatabasesReallyMakeaDifference?
ManyBIapplicationscurrentlymakeuseoflegacyRDBMSservers,butthattrendisrapidlychanging.TDWI
foundanegativetrend(specifically52%)whenitdidamajorstudyandaskedthequestioniftraditional
rowbaseddatabasesweregoingtobeusedforBIimplementations.5Thisdevelopmentisnotsurprising
becauseexperiencedBIprofessionalshaveknownforalongtimethatlegacyRDBMSsarenotdesignedfor
businessintelligenceworkloads.Theyperformverywellfortransactionprocessingsystems,butwhenit
comestodatabaseanalytics,legacydatabasevendorproductsdonotdelivereitherthespeedortheprice
tagthatthoserollingoutmodernBIapplicationsneed.

ThesameTDWIstudyfoundthatthemostimportantcomponentinabusinessintelligenceimplementation
wasthedatabaseserveritself.6

5
6

Russom,ibid.
Russom,ibid.
Copyright2010CalpontCorporation.AllRightsReserved.

WhyChooseaColumnDatabaseforBusinessIntelligence?

Nearlythreequartersofthestudysrespondentsindicatedthatitisthedatabasesoftware,whichisthe
mostcriticalpieceofaBIapplicationssuccess.Further,TDWIfoundthatnearlyhalfofthoseitpolledare
readytoreplacetheirdatabaseusedforbusinessintelligenceapplicationswithanother,moremodern
alternative.Whenaskedwhatthetechnicalreasonswereforthereplacement,thenumberoneanswerwas
theinabilityforthelegacyRDBMStoservicequeriesinthetimeneeded:

CitingarecentstudydonebyVentana,datawarehouseauthorDavidStodderbringstolightthatnearly
60%ofdatawarehouseimplementationsareexperiencingnightmarishperformanceproblemswhenthey
havetoscaleandruncomplexqueries,andnearly50%saytheyhavethesameproblemwhenitcomesto
loadingmoredata.TheendresultsareBIqueriesbeingcanceledbecausetheysimplyruntoolong.Because
Copyright2010CalpontCorporation.AllRightsReserved.

WhyChooseaColumnDatabaseforBusinessIntelligence?

ofthis,theVentanastudyfoundtheseorganizationsarenowseriouslyevaluatingcolumndatabasesto
solvetheirproblems.7

Notingthatdatavolumesaregrowingatarampantratewithinmostenterprises(withestimatesranging
from20percentperyearontheconservativesideto50percentperyearamongthelargestorganizations),
ForresterResearchsaysthattheprognosisis"withinafewyears,traditionalroworientedrelational
databasesarereallygoingtobepushingtheirlimits.TherearealotofDBAswhodontwanttohearit
becausethey'resoentrenchedinOracleandIBMDB2andMicrosoftSQLServer,butwethinkanewerais
coming."8

DocolumndatabasesreallyhavethecapabilitytodeliverwhatITexecutivesneed?Commentingonthe
technologybenefitofcolumnorienteddatabasesisGartnerGroupsdatabaseexpertDonaldFeinbergwho
says,"Ifyou'rebringingbackallthecolumns,acolumnstoredatabaseisn'tgoingtoperformanybetter
thanarowstoreDBMS,butanalyticapplicationsaretypicallylookingatallrowsandonlyafewcolumns.
WhenyouputthattypeofapplicationonacolumnstoreDBMS,itoutperformsanythingthatdoesn'ttake
acolumnstoreapproach."9

ABenchmarkComparisonbetweenInfiniDBandaLeadingRowBasedDatabase
AsanexampleofhowacolumnorienteddatabasecanoutperformalegacyRDBMS,Calpontrecently
commissionedawellknowndatawarehouseindustryexperttobenchmarktheleadingrowbased
database(ofwhichtheexperthasmanyyearsofexperienceintuningforfastperformance)against
InfiniDB,whichhasasoneofitscorefeatures,acolumnorienteddesign.TheStarSchemastyled
benchmarkwasconductedontwodifferentmachinestogaugeperformanceonbothmidandlargesized
servers.Themidsizedserverwasan8CPU,8GBRAM,14SATA7200RAID0nocacheconfiguration,and
thelargeserverwasa16CPU,16GBRAM,14SAS15KRPMRAID0with512MBcachemachine.Bothwere
running64bitCentOS5.4.Therawdatabasesizewas2TB.

Ascanbeseenonthegraphsbelow,variousconfigurationswereusedfortheleadingrowbaseddatabase,
howevernomattertheconfiguration,thecolumnorientedInfiniDBdatabaseconsistentlyanddramatically
beatthelegacydatabaseinstoragefootprint,loadtime,andqueryspeed:

DavidStodder,NineBIMegatrendsfor2009,IntelligentEnterprise,January,2009.
DougHenschen,ColumnStoreDatabasesandDWAppliances:HowtoMaketheRightChoice,InformationWeek
ResearchandReports,March2008.
9
Henschen,ibid.
8

Copyright2010CalpontCorporation.AllRightsReserved.

WhyChooseaColumnDatabaseforBusinessIntelligence?

Copyright2010CalpontCorporation.AllRightsReserved.

WhyChooseaColumnDatabaseforBusinessIntelligence?

Insummary,theInfiniDBserversavesonstoragecosts,suppliesfasteraccesstonew/incomingdata,and
runsquerymuchfasterthanitsrowbasedcompetitor.

Inadditiontoproducingoverallfasterqueryspeeds,InfiniDBalsosuppliedmuchbetterquerypredictability
intermsofquerytime.Whereastheleadingrowbaseddatabaseproducedwildlyvaryingminimumand
maximumquerytimesoverthevariousruns,InfiniDBhadafarmoretightlygroupofrunswhenitcameto
predictableresponsetimes.Thistranslatesintomuchbetterdependabilityfromabusinessstandpointin
ensuringBIreportsandqueriesmeetwhateverservicelevelagreementsareimposedfrombusinessusers.

AdditionalConsiderations
Inadditiontobetterperformance,thecolumnorientationaspectofcolumndatabasessuppliesanumber
ofusefulbenefitstothosewishingtodeployfastbusinessintelligencedatabases.

First,thereisnoneedforindexingaswithtraditionalrowbaseddatabases.Theeliminationofindexing
means:(1)lessoverallstorageisconsumedincolumndatabasesbecauseindexesinlegacyRDBMSsoften
balloonthestoragecostofadatabasetodoubleormoretheinitialdatasize;(2)dataloadspeedis
increasedbecausenoindexesneedtobemaintained;(3)adhocDMLworkspeedisincreasedbecauseno
indexupdatesareperformed;(4)noindexingdesignortuningworkisimposedonthedatabaseITstaff.

Second,thereisfarlessdesignworkforcedondatabasearchitectswhencolumndatabasesareused.The
needforcomplicatedpartitioningschemes,materializedvieworsummarytabledesigns,andothersuch

Copyright2010CalpontCorporation.AllRightsReserved.

10

WhyChooseaColumnDatabaseforBusinessIntelligence?

workiscompletelyremovedbecausecolumndatabasesneednoneofthesecomponentstoachieve
superiorqueryperformance.

TheLearningCurvewithColumnOrientedDatabases
BothITmanagementanddatabasedevelopers/administratorswillbepleasedtofindthatthelearning
curveassociatedwithmovingfromlegacy,rowbasedRDBMSstoacolumndatabaseisverysmallifnot
completelynonexistent.Unlikeotherdatabasesthatcameonthesceneinprioryearswhichrequired
eitherdifferentprogrammingparadigms(e.g.objectorienteddatabases)orlearningnewdesign
methodologiesanddatabaseaccesslanguages(e.g.OLAPdatabases),columndatabaseslookandhandle
justlikestandardrelationaldatabases.TheyusethesameANSIstandardSQLlanguage,securitymethods,
andrequirenodevelopmentparadigmchanges.

Infact,columndatabasesactuallylessentheburdenonboththedevelopmentandadministrationstaff
becausetheydoawaywiththeneedforindexingexercises,datapartitioningschemes,supplementary
objectdesigns(e.g.materializedviews),andothersimilartasks.Theeaseofusefactor,therefore,isgreater
withcolumndatabasesthanitiswithtraditionalRDBMSs.Moreover,theydonotrequiresuchspecialized
inhouseexpertisetobuildhighlyperformantsystems.

TheExpenseofImplementingColumnOrientedDatabases
AnInformationWeeksurveyfoundthatnearly40%ofITexecutivescitedexpensivesoftwarelicensingcosts
asthenumberonereasontheyareunabletorolloutalltheBIinitiativestheyrequire,andastudy
performedbyTDWIfoundthatalmost60%ofITmanagershavecitedrecenteconomicconditionsasthe
reasonforconductingareevaluationoftheirsoftwarevendorsinanefforttoreducecostsandimplement
theBIsystemstheirbusinessneeds.10

Inaddition,therecentglobalrecessionhascausedwhatmanybelievetobeapermanentchangein
mindsetawayfromhighpricedproprietaryhardwareandsoftwaretocommoditytechnologiesthatare
oftentimesfoundtobenearlytechnicallyequal,andsometimesactuallysuperiorintermsofoverall
efficiencyandperformance.TDWIfoundthatnearly60%ofitsrespondentsarenowconsideringlowercost
optionsfortheirbusinessintelligencedatabase.Reasonsforthisincludedreducedbudgets,frozenhiring,
projectsbeingputonhold,softwareacquisitiondenials,andstaffreductions.11

GeneralWaysinwhichColumnDatabasesHelpSaveMoney
Columnorienteddatabasesandthetechnologythatisbeneficiallyutilizedbythemmaycertainlysound
enticing,butthequestionneedstobeasked:howcancolumndatabaseshelpcontainITcosts?Herearea
fewgeneralwayscolumndatabaseshelpdefrayexpenses:

Mostallcolumndatabasesexploittheadvancedincommodityhardware(e.g.64bit,multi
CPU/core)anddonotrequireexpensivemachinesbeusedtoachieveexceptionalperformance.
10

PhilipRussom,NextGenerationDataWarehousePlatforms,TDWI,2009.
Russom,ibid.

11

Copyright2010CalpontCorporation.AllRightsReserved.

11

WhyChooseaColumnDatabaseforBusinessIntelligence?

ForcolumndatabasesthatsupportMPP,thepriceperformancemetric(takingcommodity
hardwareintoaccount)canbecomequiteimpressive
Themajorityofcolumndatabasesoffersomeformofdatacompressioneitherphysicalor
logicalthathelpssavestoragecosts
Inadditiontogeneraldatacompression,manycolumndatabasesdonotrequiretheuseof
indexes,materializedviews,orothersupplementarystoragestructuresthatareusedtohelp
theperformanceoftraditionalRDBMSs.Becausethesestructuresarenotused,additional
storagecostsavingsaretheendresult
Becausecolumndatabasesdonotrequireindexesorcomplicateddatabasedesignschemes
(e.g.datapartitioning,etc.)anddomuchoftheperformancedesignworkautomaticallyforthe
enduser,oftentimestheneedforhighlyexperienceddatabasestaffisnegated.Thismeansthat
inexperiencedITstaffcancreatebusinessintelligencedatabasesthatperformextremelywell
withouthavingtopossesssophisticatedknowledgeofdatawarehousedesignsand
performancetuningtechniques
Theincreasedspeedatwhichcolumndatabasesoperatetypicallymeansthatmorebusiness
intelligencereports/requestscanbedeliveredanddecisionmakerscanmorequicklyperform
theirjob,whichlessensstaffidletime

Commentingoncolumndatabasesandcostsavings,GartnersFeinbergsays,Inthetwospecificareas
wherecolumnstoreDBMSexcel(analyticsandarchivingsolutions),therecanbealargecostsavingsand,
withanalyticapplications,alargeperformancegain.

TheInfiniDBAdvantage

InfiniDBisacolumnorienteddatabasethatisdesignedtoservicetheneedsofbusinessintelligence
applicationsaswellasanyotherscenariowheremediumtoverylargeamountsofdataneedtobequeried
andfastresponsetimesaredesired.

TheapplicationsandsystemsthatwillbenefitfromInfiniDBaremany,withtherebeingthreecommonuse
caseswhereInfiniDBcanbeexpectedtoexcel:

1. Datawarehouses,datamarts,andotherbusinessintelligence(BI)datastores
2. Generalpurposereportingdatabases
3. Readintensivesegmentsofanoverallapplicationwhereread/search/lookupportionsofthe
applicationareservedbyananalytic/querydatabaseandthetransactionalsegments(e.g.order
entry,etc.)aremanagedbyatraditionalrelationaldatabase

Inadditiontobeingacolumnorienteddatabase,InfiniDBoffersanumberofotherbenefitsthattodays
businessintelligenceuserswantintheirunderlyingdatabase.Someofthecorefeaturesandbenefitsare
listedbelow.

Copyright2010CalpontCorporation.AllRightsReserved.

12

WhyChooseaColumnDatabaseforBusinessIntelligence?

ExploitsTodaysModernHardware
InfiniDBiscapableofexploitingmodern64bitcommodityhardwareandutilizingthemanyCPUsandcores
thatcomestandardontodaysservers.Becauseofitsmultithreadeddesign,InfiniDBcanscaleuponSMP
machinesandparallelizeitsoperationsacrossthemachinesresourcestoproduceveryfastqueryresponse
times.NotethatInfiniDBcanbeusedon32bitmachinesaswellifdesired.

MPPCapable
Inadditiontobeingabletoscaleup,InfiniDBcanalsoscaleoutinmassiveparallelprocessing(MPP)fashion
acrossmultiplecommodityhardwaremachinestoachievelinearincreasesinoverallperformance.Adding
inexpensivehardwaretoanInfiniDBconfigurationallowsthedatabasetoincreaseprocessingpowerso
responsetimescanmanytimesbereducedinhalfjustwiththeadditionofanewnode.Thismakesiteasy
toadjustforgrowingdatavolumes,increaseduseractivity,orthedesiretomeetbetteroverall
performancegoals.Moreover,thiscanbeaccomplishedinadynamicfashionwheretheInfiniDBsystem
doesntgodownorneedtobetakenofflinewhennewnodesareadded.

InfiniDBsmodulararchitectureallowsateamchargedwithdevelopingafastbusinessintelligencedatabase
toscalebothI/Oanduserconcurrency.NewnodescanbeaddedbothattheI/O(PerformanceModules)
anduserconnectionlayer(UserModules)toensurenobottleneckinprocessingisencountered:

InfiniDBsMPPcapabilitiesofferanotherbenefitthatofadistributedsharednothingdatacache.Ina
multiplenodeInfiniDBconfiguration,dataisdistributedamongthevariousnodesandtheirdatacaches.No
nodesharesdatawiththeother,howeverallareaccessedintheInfiniDBMPParchitecturewhendatais
readtosatisfyqueries.Inessencethen,InfiniDBcreatesonelargelogicaldatacachethatisaccessedina
distributedfashioninparallelbyallparticipatingnodes.ThisallowsInfiniDBtoliterallycachelarge
databaseswhenenoughnodesarepresentwithgenerousamountsofmemory.
Copyright2010CalpontCorporation.AllRightsReserved.

13

WhyChooseaColumnDatabaseforBusinessIntelligence?

Ofcourse,InfiniDBcanrunonastandalonemachinewithlocaldiskstoragequitewellandofferscaleup
performanceutilizingthatserversCPUsandotherresources.ButInfiniDBisnotlimitedtobeingdeployed
onsinglemachines,andiswellsuitedforMPPprocessing.

SelfManagingDatabase

GettingexceptionaldataloadandqueryperformancefromInfiniDBdoesnotrequireseasoneddatabase
expertstodesignandtunethesystem.InfiniDBtakesthedatabasedesignandadministrationheadaches
awayfrombusinessintelligenceprofessionalsthroughitsautomanagementarchitectureframework.

First,beingcolumnoriented,InfiniDBrequiresnoindexing,whichmeansdatabaseadministratorsand
developersdonthavetoworryaboutcomplexindexingschemesandthestorageoverheadsuchstructures
impose.Beingcolumnoriented,InfiniDBobviatestheneedfortraditionalindexingbecause,inessence,the
columnistheindex.

Second,InfiniDBremovestheneedformanualdatapartitioningandstorageplacement.InfiniDBusesa
formoflogicalhorizontalrangepartitioningthatdoesnotrequirespecialstorageplacementorschema
design.Usingbothvertical(columnorientation)andlogicalhorizontalrangepartitioningallowsInfiniDBto
reduceI/Oinbothdirections(columnandrow).Bothverticalandhorizontalpartitioningareautomatically
handledbytheInfiniDBdatabaseandrequirenouserintervention.

Finally,whendeployedinanMPPfashion,InfiniDBoffersautomaticI/Oandconcurrencyscalingplus
automaticfailoverattheI/Olayer.Inaddition,patchmanagementofallparticipatingInfiniDBnodesis
easedthroughanautomatedsoftwareprovisioningcapability.Whenpatchesorupgradesneedtobe
appliedtomorethanoneInfiniDBserver,anautomaticprocesstakesthesoftwarefromthefirstnode
whereanupgradeisappliedandautomaticallyupgradesallotherparticipatingnodes.

HighSpeedDataLoading

Gettingdataloadedintobusinessintelligencedatabasesasquicklyaspossibleisahighpriorityinmost
cases.InfiniDBoffersahighspeed,bulkloadutilitythatloadsdataveryquicklyintoInfiniDBtables.

Becausetherearenoindexes,loadtimesarereducedevenfurther.InlegacyRDBMSs,astandardpractice
istodropindexesontablessodataloadscanrunfasterandnewdatacanbeaddedtothesystemmore
quickly.However,therearetwoproblemswiththisapproach:(1)theindexeswillhavetoberecreated,
whichusesupCPUandmemoryresourcesonthesystem;(2)anyqueriesthatneedtoberunduringthis
timewillrunmuchslowerbecausenoindexeswillbeavailable,whichequatestofulltablescansonlarge
tablesnotsomethinglegacydatabasesareknownforhandlingwell.Buttheseissuesarenonexistentin
InfiniDBsinceindexingisunnecessary.

Inadditiontoahighspeedbulkloader,InfiniDBsupportsanumberofotherMySQLbasedloadutilities,ETL
tools,andstandardDML(insert,update,delete)commands.

Copyright2010CalpontCorporation.AllRightsReserved.

14

WhyChooseaColumnDatabaseforBusinessIntelligence?

HighDataConcurrencyandTransactionalDataProtection
InfiniDBemploysanMVCC(multiversionconcurrencycontrol)designsoreadersneverblockwritersor
dataloadsandviceversa.Aquerywillalwaysseethedataasitexistedatthetimethequerywasissued.

Inaddition,ACIDcomplianttransactionalsupportisprovidedintheInfiniDBdatabase.Transactionscan
easilybecommittedorrolledback,anddeadlockdetectionsupportisalsoprovidedtohandleconflict
resolution.Finally,InfiniDBprovidesfullcrashrecoverycapabilities.Intheeventofasystemcrash,InfiniDB
automaticallymaintainsdataintegrityanduponsystemrestart,InfiniDBsupportsrollforwardandrollback
operationstoreturnthedatabasetoaconsistentstate.

MySQLInterface
InfiniDButilizesMySQLforitsbasicfrontenduserinterface.ThisallowsanyonefamiliarwithMySQLto
becomeimmediatelyproductivewithInfiniDB.ForthosenotacquaintedwithMySQL,thelearningcurveis
minimalasMySQLsupportsalmostallANSIstandardSQLoperations.Moreover,therearemanyfreely
suppliedGUItoolsfromMySQLaswellasothervendorsthatmaybeusedtodevelopagainstand
administeranInfiniDBdatabase.

ExtensiveApplicationDevelopmentandThirdPartySoftwareSupport
InfiniDBoffersmanydifferentdriversandAPIs(ODBC,JDBC,PHP,etc.)forconnectingtothedatabase,as
wellasplentyofSQLandapplicationdevelopmentaidsthatarebuiltintotheserver.Moreover,anythird
partysoftwarebusinessintelligencetool(oranysoftwareproductforthatmatter)thatsupportsMySQL
shouldworkwellwithInfiniDB.

Conclusion
Intheend,ITorganizationsshouldconsideracolumndatabaseoveralegacyRDBMSbecauseitcanmakea
bigimpactinhowbusinessintelligencesystemsperformandthedownstreamimpactthiscanhaveinthe
timelinessofdecisionmaking.

CalpontInfiniDB

TheHighPerformance,ScalableDataPlatformthatenablesDeepAnalytics

FormoreinformationortopurchaseCalpontInfiniDB ,pleasecontact2146189500,orvisitwww.calpont.com.

Copyright2010CalpontCorporation.AllRightsReserved.

15

You might also like