calledoracle10g

ReduceCosts&IncreaseOracleDatabaseOLTPWorkloadServiceLevels:PowerEdge2950ConsolidationtoPowerEdge11thGenerationADellTechnicalWhitePaperDell|DatabaseSolutionsEngineeringBalamuruganBRaviRamappaArpil2010APrilReduceCosts&IncreaseOracleDatabaseOLTPWorkloadServiceLevelsPageiiEXECUTIVESUMMARYTheDellenterpriseportfolioisevolvingtoincorporatebetter‐performing,moreenergy‐efficient,andmorehighly‐availableproducts.
WiththeintroductionofDell'slatestserverproductline,customershaveanopportunitytoimprovetheirtotalcostofownershipbyconsolidatingdistributedlegacyenvironments.
ThisisthethirdwhitepaperinaseriesthatdiscussesserverconsolidationonDell11Gproductline.
EarlierwhitepapersthatdiscusstheDSS/OLTPworkloadconsolidationonDellPowerEdge11G2socketserversinclude:ConsolidatingDSSWorkloadsonDellPowerEdge11GServersUsingOracle11gDatabaseReplayhttp://www.
dell.
com/downloads/global/solutions/database_11g_consolidate.
pdfc=ec&l=en&s=genConsolidatingOLTPWorkloadsonDellPowerEdge11GServershttp://www.
dell.
com/downloads/global/solutions/Consolidating_OLTP_Workloads.
pdfc=us&cs=555&l=en&s=bizThiswhitepaperfocusesonOnlineTransactionProcessing(OLTP)workloadsandconsolidationonDellPowerEdge11G4/2socketservers.
DellstrivestosimplifyITinfrastructurebyprovidingmethodstoconsolidatelegacyproductionenvironmentsandreducedatacentercomplexity.
Thetoolsandproceduresdescribedinthiswhitepapercanhelpadministratorstest,compare,validate,andimplementthelatesthardwareanddatabasesolutionbundles.
DellestablishedtheseproceduresandguidelinesbasedonlabexperimentsanddatabaseworkloadsimulationsperformedbytheDellDatabaseSolutionsEngineeringteam.
Usingthetoolsandproceduresdescribedinthisdocument,customersmaynotonlyselecttheappropriatedatabasesolutionhardwareandsoftwarestack,butalsooptimizethesolutions'totalcostofownershipaccordingtothedatabaseworkloadstheychoosetorun.
Theintendedaudienceofthiswhitepaperincludesdatabaseadministrators,ITmanagers,andsystemconsultants.
ReduceCosts&IncreaseOracleDatabaseOLTPWorkloadServiceLevelsPageiiiTHISWHITEPAPERISFORINFORMATIONALPURPOSESONLY,ANDMAYCONTAINTYPOGRAPHICALERRORSANDTECHNICALINACCURACIES.
THECONTENTISPROVIDEDASIS,WITHOUTEXPRESSORIMPLIEDWARRANTIESOFANYKIND.
2010DellInc.
Allrightsreserved.
ReproductionofthismaterialinanymannerwhatsoeverwithouttheexpresswrittenpermissionofDellInc.
isstrictlyforbidden.
Formoreinformation,contactDell.
Dell,theDELLlogo,andtheDELLbadge,PowerConnect,andPowerVaultaretrademarksofDellInc.
SymantecandtheSYMANTEClogoaretrademarksorregisteredtrademarksofSymantecCorporationoritsaffiliatesintheUSandothercountries.
Microsoft,Windows,WindowsServer,andActiveDirectoryareeithertrademarksorregisteredtrademarksofMicrosoftCorporationintheUnitedStatesand/orothercountries.
OracleisaregisteredtrademarkofOracleCorporationand/oritsaffiliates.
EMCistheregisteredtrademarkofEMCCorporation.
Intel,andXeonareeithertrademarksorregisteredtrademarksofIntelCorporation.
April2010ReduceCosts&IncreaseOracleDatabaseOLTPWorkloadServiceLevelsPage1ContentsEXECUTIVESUMMARY.
iiINTRODUCTION2SYSTEMARCHITHECTURE3TESTCONFIGURATION4TESTMETHODOLOGY.
4CONSOLIDATIONFACTOR9SUMMARY.
10TablesTable1:TestConfiguration.
4FiguresFigure1:SystemArchitecture.
3Figure2:BaseConfiguration—TPSComparisonBetweenLegacyProductionandR810TestEnvironment6Figure3:BaseConfiguration-AQRTComparisonBetweenLegacyProductionandR810TestEnvironment6Figure4:BaseConfiguration-CPUTimeComparisonBetweenLegacyProductionandR810TestEnvironment7Figure5:AtLegacySaturatedUserLoad-TPSComparisonBetweenLegacyProductionandR810TestEnvironment.
8Figure6:AtLegacySaturatedUserLoad-AQRTComparisonBetweenLegacyProductionandR810TestEnvironment.
8Figure7:AtLegacySaturatedUserLoad-CPUTimeComparisonBetweenLegacyProductionandR810TestEnvironment9ReduceCosts&IncreaseOracleDatabaseOLTPWorkloadServiceLevelsPage2INTRODUCTIONServerconsolidationcanbedefinedasmaximizingtheefficiencyofcomputerserverresources,therebyminimizingtheassociatedpower/cooling,rackfootprintandlicensingcosts.
Itessentiallysolvesafundamentalproblemcalledserversprawlinwhichmultiple,under‐utilizedserverstakeupmorespaceandconsumemorepowerresourcesthantheworkloadrequirementindicates.
OLTPdatabasesystemstypicallyservicehundredsorthousandsofconcurrentusers.
Anexampleofthistypeofsystemisatravelreservationsystemwherealargenumberofcustomersandagentsperformonlinetravelreservations,orcheckavailableflightsandschedules.
TheOLTPdatabasetransactionsperformedbythesethousandsofconcurrentusersgettranslatedintotensofthousandsofI/OrequeststothebackendstoragesubsystemdependingonthenatureoftheseOLTPtransactions.
ThedatabasehostCPUsmayonlybeefficientlyusedifthebackendstoragesubsystemisconfiguredwithasufficientnumberofdiskstohandlethelargenumberofI/Orequests.
InthecaseofanOracledatabasehostCPUsexhibitlargeIOWAITtimesinsteadofdoingusefulwork.
Inthisscenario,consolidating,upgrading,ormigratingtoafasterdatabaseserver,orscalingthenumberofCPUsormemorydoesnothelp.
ThecorrectapproachistoappropriatelyscalethebackenddisksubsystemtohandletheI/Orequests,andthenmovetothenextstageofCPUandmemorysizingasdiscussedlaterinthiswhitepaper.
TheobjectiveofthiswhitepaperistoidentifytheconsolidationfactorforanOracledatabaserunningOLTPworkloadsonlegacy9thGenerationPowerEdge29502U2sockettothenew11GPowerEdgeR8102U4/2socketservers.
AnenterprisedatabasesystemmayberunningDSS,OLTP,oramixedworkload.
TheOLTPworkloadstypicallysendthousandsofsmallI/Orequestsfromthedatabaseserverstothebackendstoragesubsystem.
ThelargeamountofI/OrequestscharacteristicoftheOLTPworkload,meansthatthebackendstoragesubsystemmusthavesufficientnumberofdiskstohandletheI/Orequestscomingfromthehosts.
Consideratwo‐nodeOracleRACdatabasehostedontwoninth‐generation(9G)PowerEdge2950dual‐socket,dual‐core,orquad‐coreserversrunningOracle10gRelease2.
Dellrecentlyannouncedtheavailabilityofitseleventh-generation(11G)serverproductlineequippedwiththechipsetthatisdesignedtosupportIntelXeon7500series4/6/8coreprocessors,QuickPathInterconnect,DDR3memorytechnologyandPCIExpressgeneration2.
Apotentialreplacementfor9G2UDellserversisthe11G2UDellPowerEdgeR810server.
TheR810supportsfour‐socket,eight‐coreprocessors,twodifferenttypesofenergyefficientCPUs,andhasahighlyefficientoverallarchitecture.
Amulti‐nodeOracleRACclusteronlegacysystemswith2socketdualcoreprocessorscanbereplacedbyanOracleRACclusterconsistingoffewernodesofPowerEdge11Gwith4socketeightcoreprocessors,andstillprocesstheOLTPworkloadfasterwithlesspowerconsumptionandlowerOracleRAClicensingcost.
ThesavingsinRAClicensingmaybeusedtoefficientlyconfigureandscalethebackendstoragesystemwithenoughI/OmodulesanddiskstoremovetheI/ObottlenecksthatarealmostalwaysanissueinanOLTPenvironment.
Also,basedontheresultsofthisstudy,onemaydeterminehowmanydistributedstandalonelegacyenvironmentsrunningOLTPworkloadscanbeconsolidatedonasingleOracleRACsolutionrunningonDellR810servers.
ReduceCosts&IncreaseOracleDatabaseOLTPWorkloadServiceLevelsPage3SYSTEMARCHITHECTUREFigure1:SystemArchitectureAsshownintheabovefigurethelegacyproductionenvironmentconsistsof2nodeOracle10gR2RACrunningon9G2U2-socketPowerEdge2950IIIserversandthetestenvironmentisansinglenodeORACLE11gR2RACrunningonan11G2U4/2socketPowerEdgeR810server.
Note:ItshouldbenotedthattheintentofthispaperisnottorecommendconvertingaRACclustertoasinglenodesetup.
ThetestsetupwasdesignedtocomparethehostCPUbehaviorforconsolidatingdatabaseworkloadatthesametimeensuringthatthenumberofcoresaresameinboththesetups.
TosimulateOracleRACoverheadinsinglenodeconfiguration(withR810server),Oracle11gR2databasewasconfiguredwithOracle11gR2gridinfrastructure.
ReduceCosts&IncreaseOracleDatabaseOLTPWorkloadServiceLevelsPage4TESTCONFIGURATIONTable1describesthecompletesoftwareandhardwareconfigurationthatwasusedthroughouttestingonboththesimulatedlegacyproductionenvironmentandthe11Gtestenvironment.
Table1:TestConfigurationComponentLegacyProductionEnvironmentR810TestEnvironmentSystemsTwoPowerEdge2950III,2UserversOnePowerEdgeR8102U4/2socketserverProcessorsTwoIntelXeonX5460,3.
16GHzquadcorepernodeCache:L2=2x4MperCPUOneIntelXeonX7560,2.
26GHzeightcoreCache:L2=8x256kL3=24MMemory32GBDDR2pernode(64Gbtotal)64GBDDR3InternaldisksTwo73GB2.
5"SASpernodeTwo73GB2.
5"SASNetworkTwoBroadcomNetXtremeIIBCM5708GigabitEthernetFourBroadcomNetXtremeIIBCM5709GigabitEthernetExternalstorageDell|EMCCX4‐480with146GBFibreChanneldisksDell|EMCCX4‐480with146GBFibreChanneldisksHBAOneQLE2462pernodeOneQLE2462OSEnterpriseLinux4.
6EnterpriseLinux5.
4OraclesoftwareOracle10gR210.
2.
0.
4FileSystem:ASMDiskgroups:DATABASE,DATAsga_target=1600Mpga_target=800MOracle11gR211.
2.
0.
1.
0FileSystem:ASMDiskgroups:DATABASE,DATAmemory_target=2400MWorkloadQuestBenchmarkFactoryTPCCworkloadScalefactor:3000Userconnections:200‐5000QuestBenchmarkFactoryTPCCworkloadScalefactor:3000Userconnections:200‐5000TESTMETHODOLOGYDell'sDatabaseSolutionengineersusedQuestSoftwareBenchmarkFactoryTPCC,aload-generatingutilitythatsimulatesOLTPusersandtransactionsonadatabaseforagivennumberofusers.
TheTPCCworkloadprovidedbytheBenchmarkFactoryschemasimulatesanorderentrysystemconsistingofmultiplewarehouses,withdatapopulatedintableswithrowsaccordingtothescalefactordefinedduringtablecreation.
ThemostcommonlyusedmetricsforanOLTPenvironmentaretransactionpersecond(TPS)andaveragequeryresponsetime(AQRT).
TheAQRTofanOLTPdatabaseenvironmentmaybedescribedastheaveragetimeittakesforanOLTPtransactiontocompleteanddelivertheresultsofthetransactiontotheenduser.
TheAQRTisthemostimportantfactorwhenitcomestofulfillingend-userrequirements,anditestablishestheperformancecriteriaforanOLTPdatabase.
The2-secondsresponsetimemetricwaschosenasthebasisforourServiceLevelAgreement(SLA)whichwasmaintainedthroughoutthetestingOurinitialgoalwastostressthelegacysystemPowerEdge2950IIItodeterminetheoptimalperformanceintermsofuserloadandTPS,ensuringthatthereisnobottleneckfromastorageperspectiveaswellasthatofthehostmemory.
ThelegacydatabasewasconfiguredwithascalefactorReduceCosts&IncreaseOracleDatabaseOLTPWorkloadServiceLevelsPage5of3000thatcreatedacoupleoftableswithmillionsofrows.
Thetotaldatabasesizethatresultedwiththisscalefactorwasaround290GB.
Initiallythebackendstoragesubsystem,consistingofaDell/EMCCX4‐480storagearray,wasconfiguredwithten15KRPM146GBdisksinaRAID10configuration.
Oncepopulated,westartedwith200concurrentusersandincreasedtheuserloadto5000inincrementsof200,randomlyrunningtransactionsagainstthelegacydatabasewhilemakingsurethattheAQRTalwaysstayedbelow2seconds.
Thetestmethodologyusedisasfollows:1.
Tosimulatethelegacyproductionenvironment,atwo‐nodeOracle10gR2RACclustercomprisedofaPowerEdge2950IIIwithquad‐core,dual‐socket3.
16GHzCPU,connectedtoaDellEMCCX4-480storagesystemconfiguredwitha100GBLUNforthedatabaseSYSTEM,a400GBLUNforDATAASMdiskgroups,anda2GBLUNforthevotingandOracleClusterRegistry(OCR)partitions.
2.
UsingtheQuestSoftwareBenchmarkFactoryTPCCworkloadpopulatedthetestdatawithascalefactorof3000intothelegacyserversimulatedproductionenvironment.
3.
Afterdatapopulation,weusedtheOracleDataPumptoexportdataattheschemalevelandavoidadatareloadforeachtestiteration.
expdpsystem/oracle@racdb1SCHEMAS=questCONTENT=alldirectory=export;4.
Westartedthefirsttestiterationwithabaseconfigurationof10disksfortheDATAASMdiskgroupand200userloadtoestablishthesaturationpointofthelegacyproductionenvironment.
Wethenincreasedtheuserloadin200userincrementswhileconstantlymonitoringtheAQRT.
OncetheAQRTexceeded2seconds,thetestwasstopped.
5.
AftereachiterationweconductedahostCPUtimeanalysistodeterminethelimitingfactorforhostperformance.
6.
Oncetheback‐endspindlesweresaturated,theystartexhibitinglargeI/Olatency.
ThisresultedinlargeIOWAITatthehostCPUandalargeAQRT.
ToreducetheIOWAITatthehostCPUthenumberofspindleswereincreasedby10disksfortheDATAASMdiskgroupforthenextiterationperformed.
TheabovemethodologywascontinuedtillthehostCPUwasoptimallyutilizedwithasmallerIOWAITtime.
Atthesametime,wemonitoredwhetherornotwewereabletolowertheaveragequeryresponsebelow2secondswithahigheruserloadcomparedtotheearlieriteration.
7.
Tosimulateourtestenvironment,weconfiguredOracle11gR2singlenodeRACcomprisingofPowerEdgeR810serverpopulatedwithtwoSocketshavingeightcoreseach,tomatchthetotalCPUcorestothatofthelegacyproductionenvironment.
UsingtheQuestSoftwareBenchmarkFactory,wepopulatedthetestdatawiththesameTPCCscalefactorthatwasusedforthelegacyproductionenvironment.
8.
Thetestiterations,similartothelegacyproductionenvironment,werecarriedoutwithinanR810testenvironmentuntilwematchedtheuserloadtothemaximumuserloadsupportedonthelegacyproductionenvironment,withanSLAof2secondsAQRT.
Figures2and3belowcomparetransactionspersecondandAQRTbetweenthelegacyproductionandtheR810testenvironmentusingthebaseconfigurationof10disksRAID10ASMdiskgroup.
ReduceCosts&IncreaseOracleDatabaseOLTPWorkloadServiceLevelsPage6Figure2:BaseConfiguration—TPSComparisonBetweenLegacyProductionandR810TestEnvironmentFigure3:BaseConfiguration-AQRTComparisonBetweenLegacyProductionandR810TestEnvironment01020304050607080902004006008001000120014001600TPSUserLoadTPSlegacy10disksTPSR81010disks00.
511.
522.
533.
544.
520040060080010001200140016001800AQRT(sec)UserLoadAvgResponseTimelegacy10disksAvgResponseTimeR81010disksReduceCosts&IncreaseOracleDatabaseOLTPWorkloadServiceLevelsPage7InFigures2and3,itisobservedthatthelegacyproductionenvironmentexhibitssimilarperformanceintermsoftransactionspersecondandtheAQRTwhencomparedtotheR810testenvironment.
Donotbemisledbytheseresults.
FurtheranalysisofthehostCPUtimeintermsofUSERtimeandIOWAITtimesrevealedthatthelegacyproductionenvironmentexhibitedahigherUSERtimetoIOWAITtimeratioascomparedtotheR810testenvironmentasshowninFigure4.
Figure4:BaseConfiguration-CPUTimeComparisonBetweenLegacyProductionandR810TestEnvironmentTheabovechartrevealsaveryinterestingfact:incomparisontothelegacyproductionenvironmenttheR810testenvironment,havingthefasterCPUandoverallmoreefficientdesign,wasabletohandletheOLTPworkloadmuchfaster,andexhibitedalowUSERtoIOWAITtimeratioaswell(0.
45forlegacyvs.
0.
11forR810at1400userload).
Sinceboththeenvironmentshadanidenticalstorageconfiguration,thereasonforhigherIOWAITandlowerUSERCPUtimeontheR810testenvironmentwasduetothefasterprocessingpoweravailableonthatenvironmentascomparedtothelegacyproductionenvironment.
Overall,Figure4revealsthatinordertotakeadvantageofthefasterprocessingpoweroftheR810Gtestenvironment,weneedtoremovetheI/ObottlenecktoreducetheIOWAITtime.
Thisresultledtofurthertestsandanalysis,andwedecidedtoverifyourconclusionsbytryingtoalleviatesomeoftheI/Obottlenecksfrombothourlegacyproductionandthe11Gtestenvironmentbyincreasingthespindlecountinanincrementof10disksforourDATAdiskgroup.
ForlegacyproductionenvironmentwecontinuedtheiterationsbyincreasingthenumberofdisksuntilwereachedtheminimalIOWAITonhostCPU.
ForthisCPUsaturationpoint,wecapturedthemaximumuserloadsupported(withAQRTof2seconds).
Wetermedthisuserloadas'legacysaturationuserload'.
FornewR810testenvironment,weperformedsimilariterationswithincreasingthenumberofdisksuntilwereachedthe'legacysaturationuserload'.
Figures5and6comparethetestresultsfortheR810testenvironmentandthelegacyproductionenvironmentatthe'legacysaturationuserload'.
01020304050607080averageiowaitaverageusertimeaveragesystemtimeCPUUtilizationLegacy:10DisksR810:10DisksReduceCosts&IncreaseOracleDatabaseOLTPWorkloadServiceLevelsPage8Figure5:AtLegacySaturatedUserLoad-TPSComparisonBetweenLegacyProductionandR810TestEnvironmentFigure6:AtLegacySaturatedUserLoad-AQRTComparisonBetweenLegacyProductionandR810TestEnvironmentAsseeninFigures5and6,atthelegacyproductionenvironmentssaturateduserloadtheTPSandtheAQRTonboththeenvironmentaresimilar.
05010015020025020040060080010001200140016001800200022002400260028003000320034003600380040004200TPSUserLoadAtLegacysaturateduserloadTPSlegacyTPSR81001234567891020040060080010001200140016001800200022002400260028003000320034003600380040004200AQRT(sec)UserLoadAtLegacysaturateduserloadTPSlegacyTPSR810ReduceCosts&IncreaseOracleDatabaseOLTPWorkloadServiceLevelsPage9Figure7:AtLegacySaturatedUserLoad-CPUTimeComparisonBetweenLegacyProductionandR810TestEnvironmentAnalysisoftheCPUtimeinFigure7revealedthattheCPUusertimeforthelegacyproductionenvironmentincreaseddrasticallyandtheCPUwasoptimallyutilizedforproductivework,whichwasobviousastheCPUIOWAITtimenearedzero.
Thusweconcludedthatthelegacyproductionenvironmentwillnotscaleuponfurtheradditionofdisksandhostmemory.
AlsoitisobservedthattheaverageCPUusertimefortheR810testenvironmentisabout40%ofthetotalCPUtime,whereasonthelegacyproductionenvironmentitisabout70%.
AsCPUIOWAITtimeontheR810testenvironmentisabout38%ofthetotalCPUtime,thisenvironmentcanbescaledfurtherbyreducingtheCPUIOWAITtimeandusingthesamefordoingproductivework.
Thiscanbeachievedbyfurtherincreasingnumberofdisksatthebackend.
CONSOLIDATIONFACTORBasedonabovetestresults,onecanconcludethatthesinglenodeOracle11gR2RACrunningon11GPowerEdgeR8104/2socketserverpopulatedwith2sockets(with8coreprocessors)wasabletohandletheOLTPworkloadofatwo-nodeOracleRACclusterrunningon9GPowerEdge2950IIIserverspopulatedwithall2sockets(withquadcoreprocessors).
Thuswecanachieveaconsolidationfactorof4whenwefullypopulateR810server.
Usingthisconsolidationfactor,wecanconsolidateanOracleRACclusterwithmanynodestofewernodes.
Forexample,ifwepopulateallfoursocketsofR810serverina2nodeOracleRACsetup,thenitcanaccomplishtheOLTPworkloadofaeight-nodeOracleRACrunningonPowerEdge2950IIIservers,providedboththeenvironmentsareconfiguredwithsufficienthostmemoryandI/Odisksubsystems.
010203040506070averageiowaitaverageusertimeaveragesystemtimeCPUUtilizationAtLegacysaturateduserloadLegacyR810ReduceCosts&IncreaseOracleDatabaseOLTPWorkloadServiceLevelsPage10AlsoasseenfromtheCPUtimeanalysisgraphs(Figure7),ontheR810serverstheCPUutilizationforproductiveworkwasonlyabout40%.
ThustherewasenoughheadroomintheCPUforscalingupfurtherintermsofuserloadsupportedfortheassumedAQRTof2secondsbyreducingtheCPUIOWAITtime.
WemaythenconcludethatiftheCPUIOWAITtimeontheR810systemisbroughtdowntoalmostnil,thenwecanhaveahigherconsolidationfactorwhichwillbeashighas7.
SUMMARYDatabasesystemsrunningOnlineTransactionProcessingworkloadsrequiretheoptimalbackendstoragedisklayoutanddiskquantitiestoefficientlyservicealargeconcurrentuserpopulation.
ThelegacyserversrunningthesetypesofworkloadshavesufferedinefficientCPUresourceusageduetothearchitecturallimitations.
Thus,onlyalimitednumberofdisksormemorycouldbeservicedbyaCPUcoreinasystem.
InthiswhitepaperwedemonstratedthatPowerEdge11GserversequippedwithXeon7500SerieschipsetsforI/Oandprocessorinterfacingremovethebottlenecksandprovideanidealplatformtoconsolidatelegacydatabaseenvironments.
TheR810chipsetisdesignedtosupportIntel'sXeon5700seriesprocessorfamily,QuickPathInterconnect,DDR3memorytechnology,andPCIExpressGeneration2.
Thisstudyalsodemonstratedthat11Gserversofferlargeperformancegainswhencomparedtooldergenerationservers.
ThedatabasesystemsrunningonPowerEdge11Gserversexhibitbetterscalabilitywhenadditionalresources,suchasdisksandmemory,areadded.
CustomersrunningOracle9ior10gRACenvironmentsonlegacyserversandstoragewillbenefitfromthefindingsinthetestmethodologiesoutlinedinthiswhitepapertoconsolidatepower‐hungryRACnodesintofewer,faster,moreenergyefficientnodes.
Asdiscussedinearliersection,customerscanexpectaconsolidationfactorofatleast4(whichcangoupto7)dependingondifferentdatabaseusagepattern.
TheresultinglegacyRACnodeconsolidationcanalsodrivedownOraclelicensingcosts,resultinginsavingsthatyoucanusetoincreasebackendstorageresourcestoimproveAQRT,implementdisasterrecoverysitesandadditionalRACtest‐bedsitesforapplicationdevelopmentandtesting.
ThereducednumberofnodesdoesnotcompromiseperformancewhenpairedwithPowerEdge11Gservers.
Theresultislessclusteroverhead,simplifiedmanagement,andpositivemovementtowardanobjectiveofsimplifyingITandreducingcomplexityindatacenters.