integersisoftwaresandra

MyFirstServer:TheDellPowerEdgeT110IIComparedtoHPEntryLevelServersADellTechnicalWhitePaperDonHoffmanMyFirstServer:TheDellPowerEdgeT110IIandHPEntryLevelServerComparisonTHISWHITEPAPERISFORINFORMATIONALPURPOSESONLY,ANDMAYCONTAINTYPOGRAPHICALERRORSANDTECHNICALINACCURACIES.
THECONTENTISPROVIDEDASIS,WITHOUTEXPRESSORIMPLIEDWARRANTIESOFANYKIND.
2011DellInc.
Allrightsreserved.
ReproductionofthismaterialinanymannerwhatsoeverwithouttheexpresswrittenpermissionofDellInc.
isstrictlyforbidden.
Formoreinformation,contactDell.
Dell,theDELLlogo,andtheDELLbadge,andPowerEdgearetrademarksofDellInc.
Thecomparisonpresentedinthispaperisbasedontheenterprise-classserversconfiguredassimilarlyaspossibleandcurrentlyshippingbyDellandHP.
SPECandthebenchmarknamesSPECjbbandSPECpower_ssjaretrademarksoftheStandardPerformanceEvaluationCorporation.
ForthelatestSPECpower_ssj2008benchmarkresults,visithttp://www.
spec.
org/power_ssj2008/results/power_ssj2008.
html.
ForthelatestSPECjbb2005benchmarkresultsvisithttp://www.
spec.
org/jbb2005/results/jbb2005.
html.
SiSoftwareandthebenchmarknameSandraaretrademarksoftheSiSoftwareCompany,UK.
ForthelatestSiSoftSandra2011benchmarkresultsandreviews,visithttp://www.
sisoftware.
net/d=reviews&f=reviews_2011&l=en&a=.
Othertrademarksandtradenamesmaybeusedinthisdocumenttorefertoeithertheentitiesclaimingthemarksandnamesortheirproducts.
DellInc.
disclaimsanyproprietaryinterestintrademarksandtradenamesotherthanitsown.
June2011MyFirstServer:TheDellPowerEdgeT110IIandHPEntryLevelServerComparisonContentsExecutiveSummary5Introduction5KeyFindings5ProcessorPerformance5CryptographicBandwidth(SecurityEncryption)5PerformanceperWatt5StorageSubsystemPerformance6Methodology7Comparison1:ProcessorPerformance8Results9Comparison1:ProcessorPerformance,Part210Results10Comparison2:CryptographicBandwidth(SecurityEncryption)12Results12Comparison3:PerformanceperWatt13Results13Comparison4:StoragePerformance.
14Results14Summary15AppendixA—TestMethodology.
16SPECpower_ssj2008Standard.
16SPECpower_ssj2008BIOSSettings16SPECpower_ssj2008OSTuning17SPECpower_ssj2008Configuration.
17SPECjbb2005BIOSSettings.
18SPECjbb2005OSTuning.
18SPECjbb2005Configuration19Sandra2011SP2bBIOSSettings.
20Sandra2011OSTuning20AppendixB—ServerHardwareConfigurationInformation.
21AppendixC—ServerFirmwareandDrivers22AppendixD—Comparison1-4DetailedResults23MyFirstServer:TheDellPowerEdgeT110IIandHPEntryLevelServerComparisonFigure1:Comparison1SiSoftSandra2011ProcessorPerformance.
9Figure2:Comparison1SPECjbb2005ProcessorPerformance.
10Figure3:Comparison2CryptographicBandwidth(SystemSecurity).
.
.
.
Error!
Bookmarknotdefined.
Figure4:Comparison3SPECpower_ssj2008ProcessingEfficiency.
.
.
.
.
.
.
Error!
Bookmarknotdefined.
Figure5:Comparison4Sandra2011StorageBandwidth.
Error!
Bookmarknotdefined.
Figure6:SPECjbb2005Results–DellPowerEdgeT110II.
22Figure7:SPECjbb2005Results–HPProliantML110G623Figure8:SPECjbb2005Results–HPProliantMicroserver.
24Figure9:SPECjbb2005Results–HPCompaq6005ProBusinessPC25Figure10:SPECpower_ssj2008Results–DellPowerEdgeT110II26Figure11:SPECpower_ssj2008Results–HPProliantML110G6.
27Figure12:SPECpower_ssj2008Results–HPProliantMicroserver23Figure13:SPECpower_ssj2008Results–HPCompaq6005ProBusinessPC23Table1:HighPerformanceConfiguration–Comparisons1,3,and48Table2:SPECpower_ssj2008ComparisonBIOSConfiguration.
10Table3:Sandra2011ComparisonBIOSConfiguration.
Error!
Bookmarknotdefined.
Table4:ServerHardwareConfigurationInformation20Table5:ServerFirmwareandDrivers21MyFirstServer:TheDellPowerEdgeT110IIandHPEntryLevelServerComparisonExecutiveSummaryIntroductionDellInc.
(Dell)commissioneditsSystemPerformanceAnalysisteamtocompareDellandHPsingle-socketserveroptionsforcustomerslookingtopurchasetheirfirstserver.
TheDellPowerEdgeT110II,HPProLiantML110G6,HPProliantMicroServer,andHPCompaq6005ProBusinessPCwerecompared.
Whilenotatrue―server‖,theHPCompaq6005ProBusinessPCwasincludedinthisstudytoshowthebenefitsofpurchasinganactualserver,insteadofusingaconverteddesktopPC.
Usingtheindustry-standardbenchmarksSPECpower_ssj2008,SPECjbb2005,andSiSoftwareSandra2011,theseserverswereratedaccordingtoperformance,performance/watt,andstoragebandwidth.
Theserverswereinstalledwiththefastestavailableprocessorandstoragesolutioneachmodelwouldsupporttoreflectthebestpossiblesolutionacustomercouldpurchase.
ThesehighperformanceconfigurationsweretestedwiththefullSandra2011SP2bbenchmarksuiteandtheSPECjbb2005benchmark.
ThesystemswerethenconfiguredwithasingleSATAdrivetoproduceabestcaseSPECpower_ssj2008score.
TheresultsshowedtheDellPowerEdgeT110IIhadthehighestperformancescoresinallbenchmarkcategories,includingprocessorandstorageperformance,cryptographicbandwidth,andperformance/watt(energyefficiency).
KeyFindingsKeyfindingsfromthestudyforpowerandperformancearesummarizedbelow.
ProcessorPerformanceTheDellPowerEdgeT110IIprovidedhigherrawprocessingperformancethaneachoftheHPserversinallbenchmarkcomparisons.
TheDellPowerEdgeT110IIoutperformedtheHPserversbyatleast33%inallSandra2011processorbenchmarks,withthehighestwinbeingbyamarginof2692%.
TheSPECjbb2005benchmarkcomparisonshowedtheDellPowerEdgeT110IItohavea57%performanceadvantageoveritsnearestcompetitorforJavaVirtualMachineprocessing.
CryptographicBandwidth(SecurityEncryption)TheSandra2011benchmarkincludesaprocessorsubtestthatmeasuresthecryptographicperformanceofthemostcommonsecurityalgorithmsinusetoday.
TheDellPowerEdgeT110IIoutperformedalloftheHPserversbyatleast3X,makingitthebestentrylevelserverfordatasecurityande-commerceapplications.
PerformanceperWattTheDellPowerEdgeT110IIachievedahigherperformancetopowerratioacrossallloadlevelsthantheHPProLiantML110G6,HPProliantMicroServer,andHPCompaq6005ProBusinessPCinalltestedconfigurations.
TheDellPowerEdgeT110IIwasmoreefficientbyarangeof57%to234%comparedtotheHPserversinoverallSPECpower_ssj2008scores.
MyFirstServer:TheDellPowerEdgeT110IIandHPEntryLevelServerComparisonStorageSubsystemPerformanceTheDellPowerEdgeT110IIoffersthehighestperformingstoragesubsystemofanyofthemodelstested.
TheDellPowerEdgeT110IIisalsotheonlyservertestedthatoffersSASdrivesinafullyhardwareacceleratedRAID0storageconfiguration.
Testmethodologyanddetailedresultsaredocumentedinthispaper.
MyFirstServer:TheDellPowerEdgeT110IIandHPEntryLevelServerComparisonMethodologySPECpower_ssj2008andSPECjbb2005areindustrystandardbenchmarkscreatedbytheStandardPerformanceEvaluationCorporation(SPEC)tomeasureaserver'spowerandperformanceacrossmultipleutilizationlevels.
Sandra2011isanindustrystandardbenchmarkcreatedbySiSoftwarethatmeasuresperformanceoftheprocessor,storage,andmemorysubsystemsindividually.
AppendixAdetailsthetestmethodologyusedbyDell,AppendicesBandCprovidedetailedconfigurationforthetests,andAppendixDprovidesdetailedreportdatathatsupportstheresultsinthispaper.
MyFirstServer:TheDellPowerEdgeT110IIandHPEntryLevelServerComparisonComparison1:ProcessorPerformanceWhencomparingservers,oneofthemostimportantvariablestoconsiderisoverallprocessorperformance.
Eachoftheserversinthisstudywasequippedwiththehighestspeedprocessoravailable,representingthehighestscoringconfigurationpossible.
Sinceeachoftheserversinthisstudyrepresentsdifferentprocessor/chipsetarchitectures,itwasnotpossibletocompareidenticalprocessorsdirectly,soeachsystemwasconfiguredformaximumperformanceforeachprocessortype.
TheconfigurationusedinComparison1issummarizedinTable1.
Thisconfigurationforeachsystem,knownastheMaximumPerformanceconfiguration,isalsoutilizedinComparisons2and4.
Table1:DetailedConfigurationforComparison1–MaximumPerformanceTheSiSoftwareSandra2011processorbenchmarkhasseveralsubteststhatareusedtoanalyzeprocessorperformance.
Thesesubtestsmeasurenumericalintegercalculations,floatingpointoperations,andrenderingofcompleximages.
Figure1showstherawperformancescoresforeachserveroneachofthefourprocessorworkloads.
1TheonlyfactoryRAIDconfigurationfortheHPCompaq6005ProBusinessPCisRAID1.
Duetothisrestriction,RAID1wasthechosenstorageconfigurationforthisstudy.
2DuetothereducednumberofcoresintheHPProliantMicroserverthesystemwastestedwithboth1and2JVMsonSPECjbbandSPECpower.
Using1JVMgavethebestscores,sothiswastheconfigurationchosen.
Comparison1DellPowerEdgeT110IIHPProliantML110G6HPProliantMicroServerHPCompaq6005ProBusinessPCSockets/FormFactor1S/Tower1S/Tower1S/Tower1S/TowerProcessorsIntelXeonE3-12703.
40GHzIntelXeonX34702.
93GHzAMDAthlonIINeoN36L1.
30GHzAMDPhenomIIX4B953.
0GHzPhysical/LogicalCores4/84/82/24/4Memory(runatmaximumspeedprocessorwillsupport)2x4GB1333MHzUDIMMs2x4GB1333MHzUDIMMs2x4GB1333MHzUDIMMsat800MHz2x4GB1333MHzUDIMMsat1200MHzHardDrives4x250GB15kRPMSAS6GbpsRAID04x250GB7200RPMSATA3GbpsRAID04x250GB7200RPMSATA3GbpsRAID02x250GB7200RPMSATA3GbpsRAID11StorageControllerDellPERCH200512MBHPSmartArrayB110iIntegratedSATARAIDControllerAMD(Xpert)RAIDControllerSoftwareConfiguration2xIBMJ9JVM2xIBMJ9JVM1xIBMJ9JVM22xIBMJ9JVMMyFirstServer:TheDellPowerEdgeT110IIandHPEntryLevelServerComparisonFigure1:Comparison1-Sandra2011SP2bProcessorPerformanceBenchmarksResultsInthismaximumperformancecomparison,theDellPowerEdgeT110IIoutperformedalloftheHPsystemsinrawprocessingpower.
ThePowerEdgeT110IIfeaturesthelatestIntelXeonProcessorE3Family,whichprovidessignificantperformanceadvantageoverthepreviousgenerationXeon3400seriesprocessorintheHPProliantML110G6.
Inthiscomparisonandtherestofthiswhitepaper,theverylowperformanceoftheHPProliantMicroserverandHPCompaq6005ProBusinessDesktopisexposed.
TheAMDAthlonIIandPhenomIIprocessorsinthesesystemsaretypicallyusedindesktopsystemsandarenotserver-classprocessors,astheconsiderableperformanceshortfallofthisstudyshows.
ThePowerEdgeT110IIistheclearwinnerinallSandra2011processorperformancebenchmarksMyFirstServer:TheDellPowerEdgeT110IIandHPEntryLevelServerComparisonComparison1:ProcessorPerformance,Part2SPECjbb2005measurestheprocessingperformanceofserversastheycomputeJavaoperations.
Thebenchmarkemulatesa3-tiersystem,themostcommontypeofserver-sideJavaapplicationtoday.
ThenumberofJVMsusedforthisbenchmarkwasoptimizedforthenumberoflogicalprocessorspresentineachserver.
AlloftheserverswereabletosupporttwoJVMs,excepttheHPProliantMicroserver,whichgotthebestresultsusingonlyoneJVMduetothelowernumberofcoresavailable.
Figure2showstherawscoresforeachservermeasuredinSPECjbb2005bops(businessoperationspersecond).
ResultsTheDellPowerEdgeT110IIagainistheclearleaderinprocessorperformance,thistimemeasuredbyJavaVirtualMachine(JVM)operationspersecond.
TheolderHPProliantML110G6isoutperformedby26%,whilethedesktopclassprocessorsintheHPProliantMicroserverandHPCompaq6005ProBusinessPCarebehindby642%and68%respectively.
Figure2:Comparison1-ProcessorPerformance–SPECjbb2005benchmarkTheDellPowerEdgeT110IIisthebestserverforJavaoperationsinthisstudyMyFirstServer:TheDellPowerEdgeT110IIandHPEntryLevelServerComparisonSPECjbb2005resultslistedbytotalSPECjbb2005bopsandSPECjbb2005bops/JVMasrequiredbySPECfairusageguidelines.
http://www.
spec.
org/fairuse.
html#JBB2005oDellPowerEdgeT110II(1chip,4cores,8threads)369,329SPECjbb2005bops,2JVMs,184,664SPECjbb2005bops/JVM.
oHPProLiantML110G6(1chip,4cores,8threads)292,782SPECjbb2005bops,2JVMs,146,391SPECjbb2005bops/JVM.
oHPProLiantMicroserver(1chip,2cores,2threads)49,690SPECjbb2005bops,1JVM,49,690SPECjbb2005bops/JVM.
oHPCompaq6005ProBusinessPC(1chip,4cores,4threads)220,277SPECjbb2005bops,2JVMs,110,138SPECjbb2005bops/JVM.
MyFirstServer:TheDellPowerEdgeT110IIandHPEntryLevelServerComparisonComparison2:CryptographicBandwidth(SecurityEncryption)Comparison2examinesthecapabilitiesoftheserversatencoding/decodingsecuretransmissionsaswellasthehashingcalculationsperformedtodetectdatacorruptionortampering.
Theseoperationsarecriticaltorunningasecurebusinessenvironmentandcancrippleasystemnotequippedforthislevelofenterpriseoperation.
Figure3:Comparison2-SiSoftwareSandra2011CryptographicBandwidth(Gb/s)ResultsAsinComparison1,theDellPowerEdgeT110IIhasasignificantadvantageinrawprocessingperformance.
DuetoadvancedAVX(AdvancedVectorExtensions)instructionsetavailableonlyinthenewIntelXeonE3architectureoftheDellPowerEdgeT110II,theperformanceadvantageovertheHPserversisevengreater.
TheHPserversarebehindincryptographicperformancebybetween60%and2692%.
ThiscomparisonshowsagainthattheolderIntelXeon3400seriesandAMDAthlonIIandPhenomII-basedHPsystemscannotkeepupwiththelevelofperformanceavailableintheDellPowerEdgeT110II.
TheDellPowerEdgeT110IIisbyfarthebestentrylevelservertestedfordatasecurityande-commerceapplicationsinthisstudy.
FormoreinformationonthenewIntelAVXinstructionset,pleasevisitthefollowingsite:http://software.
intel.
com/en-us/avx/TheDellPowerEdgeT110IIisthemostcapableserverfordatasecurityande-commercebyaratioofmorethan2to1MyFirstServer:TheDellPowerEdgeT110IIandHPEntryLevelServerComparisonComparison3:PerformanceperWattEnergycostsarerisinganditismoreimportantthaneverbeforetogetthemostprocessingpowerforyourenergydollar.
TheSPECpower_ssj2008benchmarkbytheStandardPerformanceEvaluationCorporation(SPEC)measurestheoverallperformanceperwattofeachsystemundertest.
Thebenchmarkmeasuresthesystemperformanceat100%processorutilizationaswellasthepowerdrawatthislevel.
Theprocessorutilizationissteppeddown10%atatimeuntilsystemidleisachieved.
Duringthesetimedintervalsthesystempowerusageisalsomeasured,tocalculatetheoverallperformanceperwattmetricshownbelow(ssj_ops/watt).
AppendicesAandBshowthehardwareandsoftwareconfigurationsforthePerformanceperWattcomparison.
ThesystemRAIDconfigurationswereremovedandreplacedwithsingleSATAdrivesforthebestcaseperformanceperwattforeachsystemstudied.
AsFigure4shows,theDellPowerEdgeT110IIdeliversmuchhigherprocessingpowerperwattthananyoftheHPsystemstested.
NotonlyistheDellPowerEdgeT110IIthehighestperformingserverinthisstudy,asshowninComparisons1and2,italsodoessomoreefficientlythanitscompetitors.
Figure4:Comparison3-SPECpower_ssj2008OverallPerformanceperWatt(ssj_ops/watt)ResultsDuetotheadvancedpowersavingfeaturesinthenewIntelXeonE3processorarchitectureintheDellPowerEdgeT110IIisfarmoreefficientthananyoftheHPsystems.
TheincreasedpowerusagerequiredtocompletethesameamountofworksignificantlyincreasesthetotalcostofownershipoftheHPserversinthisstudywhencomparedtotheDellPowerEdgeT110II.
Withaprocessorefficiencyofbetween57%to234%TheDellPowerEdgeT110IIistheclearwinnerinperformance/watt.
MyFirstServer:TheDellPowerEdgeT110IIandHPEntryLevelServerComparisonbetterthantheHPProliantandHPCompaq6005systems,theDellPowerEdgeT110IIprovidesmuchbettervalueoverthelifetimeoftheproduct.
Comparison4:StoragePerformanceTheabilitytoquicklyaccessbusinesscriticaldatafromtheinternalstoragesubsystemisanotherimportantaspectofserverperformance.
E-commerceandbusinesstransactionscaneasilybebottleneckedbypoorstoragesubsystemperformance.
Comparison4investigatesthemaximumstoragebandwidthofeachsysteminthisstudy.
Figure5:Comparison4Sandra2011StorageSubsystemPerformanceResultsThestoragebandwidthachievedbytheDellPowerEdgeT110IIissignificantlybetterthananyoftheHPsystemsfeaturedinthisstudy.
TheDellPowerEdgeT110IIoffersafullyhardwareaccelerated,15kSASfourdriveRAID0configuration.
TheHPProliantserversonlyofferSATAdrivesintheirhardwareacceleratedRAID0arrays,whichiswhytheyare16%-18%slower.
Aswasmentionedearlier,theHPCompaq6005ProBusinessPCdoesnotofferafactoryRAID0configuration,sothebestperformingoptionforthissystemisatwodriveSATAsoftwareacceleratedRAID1configuration.
TheRAID0option,standardonallservers,isnotpresentinadesktopclasssystemlikethe6005ProBusinessPC.
MyFirstServer:TheDellPowerEdgeT110IIandHPEntryLevelServerComparisonSummaryTheresultsofComparison1showthattheDellPowerEdgeT110IIhasmuchmorerawprocessingpowerthananyoftheHPsystemsinthisstudy.
TheDellPowerEdgeT110IIistheperformanceleaderinnumericalinteger,floatingpoint,imageprocessing,andserversideJavacalculations.
Themarginofvictorywassignificant,rangingfrom26%upto2692%.
Comparison2showedthattheadvancedAVXinstructionsfeaturedintheDellPowerEdgeT110IImakeittheidealserverforsecurebusinesstransactionsande-commerce.
TheDellPowerEdgeT110IIoffersbetween60%upto2692%betterperformancethantheHPserversforencryption/decryptionandhashingcalculations.
Whenitcomestoenergyefficiency,theDellPowerEdgeT110IIbeatstheHPserversinthisstudybyawidemargin.
Comparison3showedthattheperformance/wattadvantageoftheDellPowerEdgeT110IIis57%to234%higherthantheHPProLiantandHPCompaqDesktopmachines.
Overthelifeoftheproduct,thisincreasedpowerusagecanendupcostingasignificantamounttothebottomline.
TheDellPowerEdgeT110IIwithitsadvancedpowermanagementfeaturesisthebestchoicefortotalcostofownership.
TheabilitytoaccessbusinesscriticaldataquicklywasthegoalofComparison4.
TheSandra2011benchmarkwasusedtomeasurestoragesubsystemperformance.
TheonlyserverinthisstudytoofferahardwareRAID0configurationforSASdrives,theDellPowerEdgeT110IIwasagaintheperformanceleader,beatingtheHPProliantserversby16%and18%respectively.
TheresultsshowthattheHPCompaq6005ProBusinessPCdoesnotofferallofthefeaturesexpectedinabusinessclassserver,offeringonlyRAID1,noremotemanagementcapabilities,and351%lowerstorageperformance.
MyFirstServer:TheDellPowerEdgeT110IIandHPEntryLevelServerComparisonAppendixA—TestMethodologySPECpower_ssj2008BenchmarkSPECpower_ssj2008isanindustrystandardbenchmarkcreatedbytheStandardPerformanceEvaluationCorporation(SPEC)tomeasureaserver'spowerandperformanceacrossmultipleutilizationlevels.
SPECpower_ssj2008consistsofaServerSideJava(SSJ)workloadalongwithdatacollectionandcontrolservices.
SPECpower_ssj2008resultsportraytheserver'sperformanceinssj_ops(serversideJavaoperationspersecond)dividedbythepowerusedinwatts(ssj_ops/watt).
SPECcreatedSPEcpower_ssj2008forthosewhowanttoaccuratelymeasurethepowerconsumptionoftheirserverinrelationtotheperformancethattheserveriscapableofachievingwithssj2008workload.
SPECpower_ssj2008consistsofthreemainsoftwarecomponents:ServerSideJava(SSJ)Workload—Javadatabasethatstressestheprocessors,cachesandmemoryofthesystem,aswellassoftwareelementssuchasOSelementsandtheJavaimplementationchosentorunthebenchmark.
PowerandTemperatureDaemon(PTDaemon)—Programthatcontrolsandreportsthepoweranalyzerandtemperaturesensordata.
ControlandCollectSystem(CCS)—Javaprogramthatcoordinatesthecollectionofallthedata.
FormoreinformationonhowSPECpower_ssj008works,seehttp://www.
spec.
org/power_ssj2008/.
Allresultsdiscussedinthiswhitepaperarefrom―compliantruns‖inSPECterminology,whichmeansthatalthoughtheyhavenotbeensubmittedtoSPECforreview,Dellisallowedtodisclosethemforthepurposeofthisstudy.
AllconfigurationdetailsrequiredtoreproducetheseresultsarelistedinAppendicesA,B,andCandallresultfilesfromtherunscomparedareincludedinAppendixD.
AllserverswereconfiguredbyinstallingafreshcopyofMicrosoftWindowsServer2008FoundationsR2(ServicePack1)withtheoperatingsysteminstalledonaone-harddriveSATAconfigurationwithnoRAIDoptionsenabled.
TheonlyexceptiontothiswastheHPCompaq6005ProBusinessPC.
ItonlycomeswithWindows7EnterprisefromthefactoryanddoesnotofferaserverOSasanoption.
Duetothislimitationthissystemwasinstalledwiththeclientoperatingsystemofferedonthissystem.
TheLockPagesinMemoryoptionwasachievedbydisablingUACinthecontrolpanel,thensettingLockPagesinMemorytoEnabledforAdministrator.
ThisconfigurationisthePowerEfficientconfigurationmentionedincomparison3andshowningreaterdetailinAppendixB.
Thelatestdriverandfirmwareupdatepackagesavailabletobothserverswereinstalledatthebeginningofthisstudy.
RefertoAppendixBfordetails.
TheDellSystemPerformanceAnalysisTeamranSPECpower_ssj2008tentimesperconfigurationacrossallfourserversandchosetherunwiththehighestoverallssj_ops/wattforeachconfigurationtocompareforthisstudy.
SPECpower_ssj2008BIOSSettingsBIOSsettingsdifferedbetweenthefoursystems,sowetunedeachsystemforbest-knownSPECpower_ssj2008performanceresults.
Toimprovepowerefficiency,wechangedthememoryspeedofboththeDellPowerEdgeT110IIandHPProliantML110G6systemsto1066MHzfromthedefaultof1333MHz.
FortheAMDbasedHPProliantMicroserverandHPCompaq6005ProBusinessPCthememoryspeedwasalreadyrunningbelowmaximumspeedsothiswasnotchanged.
VirtualizationwasnotusedinthesetestssoVirtualizationsupportwasdisabledonallserversinthisstudy.
BelowisatablelistingtheSPECpower_ssj2008BIOSsettingsforeachsystemMyFirstServer:TheDellPowerEdgeT110IIandHPEntryLevelServerComparisonTable2:SPECpower_ssj2008BIOSsettingsforeachsystemSPECpower_ssj2008OSTuningToimproveJavaperformance,largepageswereenabledbyenteringControlPanel->AdministrativeTools->LocalSecurityPolicy->LocalPolicies->UserRightsAssignment->LockPagesinMemory.
AnoptionwaschangedtoaddAdministrator.
OperatingSystemPowerManagementmodeforallsolutionswaschangedtoPowerSaverandtheplanAdvancedOptionseditedtoturnofftheHardDriveafter1minute.
WeconfiguredallserverswithaseparateIPaddressonthesamesubnetasourSPECpower_ssj2008controllersystemwheretheDirector,CCS,andPTDaemoncomponentswerelocated,andconnectedbothserversdirectlytothecontrollersystemthroughNIC1fortheirrespectiveruns.
SPECpower_ssj2008ConfigurationIBMJ9JavaVirtualMachine(JVM)3wasusedforallsystems,asthisJVMprovidedthebestperformanceforSPECpower_ssj2008ofanyoftheavailablechoicesatthetimethatthisstudywasundertaken.
ThefollowingJVMoptionswereusedonallservers,astheyarethebest-knownJVMtuningsforSPECpower_ssj2008fortheIBMJ9JVMwhenrunningwithlargermemoryconfigurations:-Xms1875m-Xmx1875m-Xmn1400m-Xaggressive-Xcompressedrefs-Xgcpolicy:gencon-XlockReservation-Xnloa-XtlhPrefetch-XlpThebindingschosenweredifferentforeachsystembecauseofthedifferingcorecountsandnumberofJVMsusedforeachconfiguration.
Belowisabulletlistoftheaffinityforeachsystem'sJVMstothetotalavailablenumberoflogicalprocessors.
DellPowerEdgeT110II-start/affinity[F,F0]HPProliantML110G6-start/affinity[F,F0]HPProliantMicroserver-start/affinity[3]HPCompaq6005ProBusinessPC-start/affinity[3,C]3JVMbuild2.
4,J2RE1.
6.
0IBMJ92.
4WindowsServer2008amd64-64jvmwa6460sr5-20090519_35743MyFirstServer:TheDellPowerEdgeT110IIandHPEntryLevelServerComparisonPowerMeterConfigurationWeusedtheYokogawaWT210DigitalPowerMeterfortheactualpowermeasurementoftheservers,asthisisthemostcommonlyusedanalyzerforSPECpower_ssj2008publicationsatthetimethatthisstudywasundertaken.
SPECjbb2005SPECjbb2005isanindustrystandardbenchmarkcreatedbytheStandardPerformanceEvaluationCorporation(SPEC)tomeasureaserver'sServerSideJava(SSJ)performance.
SPECjbb2005evaluatestheperformanceofserversideJavabyemulatingathree-tierclient/serversystem(withemphasisonthemiddletier).
ThebenchmarkexercisestheimplementationsoftheJVM(JavaVirtualMachine),JIT(Just-In-Time)compiler,garbagecollection,threadsandsomeaspectsoftheoperatingsystem.
ItalsomeasurestheperformanceofCPUs,caches,memoryhierarchyandthescalabilityofsharedmemoryprocessors.
FormoreinformationonSPECjbb2005,seehttp://www.
spec.
org/jbb2005/.
Allresultsdiscussedinthiswhitepaperarefrom―compliantruns‖inSPECterminology,whichmeansthatalthoughtheyhavenotbeensubmittedtoSPECforreview,Dellisallowedtodisclosethemforthepurposeofthisstudy.
AllconfigurationdetailsrequiredtoreproducetheseresultsarelistedinAppendicesA,B,andCandallresultfilesfromtherunscomparedareincludedinAppendixD.
AllserverswereconfiguredbyinstallingafreshcopyofMicrosoftWindowsServer2008FoundationsR2(ServicePack1)withtheoperatingsysteminstalledonaRAID0configurationwhenavailable.
TheonlyexceptiontothiswastheHPCompaq6005ProBusinessPC.
ItonlycomeswithWindows7EnterprisefromthefactoryanddoesnotofferaserverOSasanoption.
Duetothislimitationthissystemwasinstalledwiththeclientoperatingsystemofferedonthissystem.
TheLockPagesinMemoryoptionwasachievedbydisablingUACinthecontrolpanel,thensettingLockPagesinMemorytoEnabledforAdministrator.
FortheHPCompaq6005ProBusinessPCRAID1wastheonlyRAIDoptionavailable.
ThisconfigurationistheMaximumPerformanceconfigurationmentionedinComparison1,Table1.
Thelatestdriverandfirmwareupdatepackagesavailabletoallserverswereinstalledatthebeginningofthisstudy.
RefertoAppendixBfordetails.
TheDellSystemPerformanceAnalysisTeamranSPECjbb2005tentimesperconfigurationacrossallfourserversandchosetherunwiththehighestSPECjbb2005bopsforeachconfigurationtocompareforthisstudy.
SPECjbb2005BIOSSettingsThesameBIOSsettingswereusedfortheSPECjbb2005testingaswasusedforSPECpower_ssj2008.
Table2showsthisconfigurationearlierinAppendixA.
TheonlyexceptionwasthatfortheIntelbasedsystemsthememoryfrequencywasadjustedbacktothedefault(maximumperformance)setting.
ThepowerplanforeachsystemwasselectedasMaximumPerformanceintheBIOStomatchtheOSsettingsbelow.
SPECjbb2005OSTuningTheoptimalsettingschosenfortheSPECpower_ssj2008comparisonarealsothebestonesforSPECjbb2005testing.
TheonlydifferenceistheOSpowerplanwassettoMaximumperformanceandtheAdvancedOptionseditedtoresettheharddrivesleeptimeto20minutes.
MyFirstServer:TheDellPowerEdgeT110IIandHPEntryLevelServerComparisonSPECjbb2005ConfigurationIBMJ9JavaVirtualMachine(JVM)4wasalsousedforallSPECjbb2005testingasthisJVMprovidedthebestperformanceforSPECjbb2005ofanyoftheavailablechoicesatthetimethatthisstudywasundertaken.
TheJVMbindingsforSPECjbbwerealsothesameastheonesusedinSPECpower_ssj2008listedabove.
SiSoftwareSandra2011Sandra2011isanindustrystandardbenchmarkcreatedbytheSiSoftware.
Thisbenchmarkwascreatedtomeasuretheperformanceofeachsubsysteminacomputer.
Thisdatacanbeanalyzedindividuallytoinvestigatespecificsubsystemperformanceorasawholetobenchmarktotalsystemperformance.
Thisstudychoseasubsetofthetotalmetricsavailable,onesthatwouldberelevanttosmallbusinessservercustomers.
Belowisalistofeachbenchmarkandashortdescriptionofwhatitmeasures:AggregateArithmeticPerformance–measuredinGigaOperationsPerSecond(GOPS),thismetricisaratingderivedfromtheIntegerandFloatingPointprocessortestswithinSandra2011.
DhrystoneiSSE4.
2–measuredinGigaIntegeroperationsPerSecond(GIPS),thistestmeasuresthecapabilityofaprocessortocalculateintegerbasedoperations.
ExamplesofthisarecomplexmatrixmultiplicationandMonteCarlosimulations.
WhetstoneiSSE3(GigaFLoatingpointOperationsPerSecond)–thisbenchmarkmeasuresthefloating-pointcomputationalperformanceofasystem.
Thistypeofcalculationismostcommoningraphicrenderingandscientificanalysis.
ProcessorMultimedia(MegaPixelsperSecond)–thefinalSandra2011processortest,thismeasuresthecalculationspeedachievedwhenrenderingaseriesofcomplexfractalimages.
CryptographicBandwidth(Gb/s)–thisisaspecializedtestthatmeasuresthetotalthroughputofasystemwhencomputingthemostcommonsecurityencryptionanddecryptionfunctions.
Also,includedinthistestarethefilehashingcalculationsnecessarytodetectfilecorruptionandtampering.
StoragePerformance(MB/s)–thistestmeasuresthestoragesubsystemperformance.
Usingsequentialreadsandwrites,themaximumlevelofdatatransferisachievedandrecorded.
FormoreinformationonSiSoftwareSandra2011,seehttp://www.
sisoftware.
net/Sandra2011SP2bwaspurchaseddirectlyfromSiSoftwareandinstalledfromtheUSBkeyprovided.
Allbenchmarkoptionsremainedsettothedefaultconfigurationandresultedinscorescompliantwiththebenchmarkrunrules.
AllserverswereconfiguredbyinstallingafreshcopyofMicrosoftWindowsServer2008FoundationsR2(ServicePack1)withtheoperatingsysteminstalledonaRAID0configurationwhenavailable.
TheonlyexceptiontothiswastheHPCompaq6005ProBusinessPC.
ItonlycomeswithWindows7EnterprisefromthefactoryanddoesnotofferaserverOSasanoption.
Duetothislimitationthissystemwasinstalledwiththeclientoperatingsystemofferedonthissystem.
TheLockPagesinMemoryoptionwasachievedbydisablingUACinthecontrolpanel,thensettingLockPagesinMemorytoEnabledforAdministrator.
FortheHPCompaq6005ProBusinessPCRAID1wastheonlyRAIDoptionavailable.
Also,theHPCompaq6005ProBusinessPConlyoffersWindows7Enterprisefromthefactory,sothiswastheconfigurationtested.
ThisconfigurationistheMaximumPerformanceconfigurationmentionedinComparison1,Table1.
Thelatestdriverandfirmwareupdatepackagesavailabletoallserverswereinstalledatthebeginningofthisstudy.
RefertoAppendixBfordetails.
4JVMbuild2.
4,J2RE1.
6.
0IBMJ92.
4WindowsServer2008amd64-64jvmwa6460sr5-20090519_35743MyFirstServer:TheDellPowerEdgeT110IIandHPEntryLevelServerComparisonSandra2011BIOSSettingsTheBIOSsettingsfortheSandra2011runsdifferedfromtheSPECrunsinthatallhardwareprefetcherswerechangedbacktotheirdefaultsettingofenabled.
Table3belowshowsthisconfiguration.
ThepowerplanforeachsystemwasselectedasMaximumPerformanceintheBIOStomatchtheOSsettingsbelow.
Table3:Sandra2011BIOSsettingsforeachsystemSandra2011OSTuningTheoptimalsettingsforSandra2011aretosettheOSpowerplantoMaximumperformanceandtheAdvancedOptionseditedtoresettheharddrivesleeptimeto20minutes.
NootherspecialconfigurationoftheoperatingsystemorthebenchmarkitselfisrequiredtosetupandrunSandra2011.
TheoptionforrunningallSandra2011testswasselectedandthebenchmarkexecutedthechosenscriptwithoutanyfurtheruserinteraction.
MyFirstServer:TheDellPowerEdgeT110IIandHPEntryLevelServerComparisonAppendixB—ServerHardwareConfigurationInformationTable4:ServerHardwareConfigurationInformationDellPowerEdgeT110IIHPProLiantML110G6HPProLiantMicroserverHPCompaq6005ProBusinessPCMemoryModulesTotalRAMinsystem(GB)8888VendorandmodelnumberHynixGMT251U7BFR8AHynixGMT251U7BFR8AHynixGMT251U7BFR8ASamsungM391B5273CH0-YH9TypePC3L-10600EPC3L-10600EPC3L-10600EPC3L-10600ESpeed(MHz)1333133313331333Speedinsystemastested1333and1066MHz1333and1066MHz800MHz1200MHzTiming/latencyCAS9CAS9CAS9CAS9NumberofRAMmodules2x4GB2x4GB2x4GB2x4GBRankorganizationDualRankx8DualRankx8DualRankx8DualRankx8HardDiskVendorandmodelnumberDellST3450857SSSeagateST3500418ASSeagateST3500418ASSeagateST3500418ASNumberofdisksinsystem4442Size(GB)500250250500RPM15,000720072007200TypeSAS6GbpsSATA3GbpsSATA3GbpsSATA3GbpsRAIDTypeRAID0RAID0RAID0RAID1ControllerDellPERCH200HPSmartArrayB110iEmbeddedAMDControllerAMD(Xpert)RAIDControllerEmbeddedOperatingSystemNameMicrosoftWindowsServerFoundations2008R2EnterpriseSP1MicrosoftWindowsServerFoundations2008R2EnterpriseSP1MicrosoftWindowsServerFoundations2008R2EnterpriseSP1MicrosoftWindows7EnterpriseBuildnumberFilesystemNTFSNTFSNTFSNTFSLanguageEnglishEnglishEnglishEnglishNetworkAdapterVendorandmodelnumberBroadcomBCM5772NetXtremeIIBroadcomNC107iNetXtremeBroadcomNC107iNetXtremeBroadcomBCM5761NetXtremeTypeIntegratedIntegratedIntegratedIntegratedMyFirstServer:TheDellPowerEdgeT110IIandHPEntryLevelServerComparisonAppendixC—ServerFirmwareandDriversTable5:ServerFirmwareandDriversDriver/FirmwareVersionsDellPowerEdgeT110IIHPProLiantML110G6HPProliantMicroserverHPCompaq6005ProBusinessPCSystemBIOS1.
0.
32010.
12.
15(Released3/1/2011)2011.
01.
17(A)(Released2/28/2011)1.
12Rev.
A(Release4/1/2011)NetworkFirmware6.
2.
142.
1.
5.
9(B)6Oct20102.
1.
5.
9(B)6Oct20102.
0Rev.
(A)3May2010NetworkDrivers16.
2.
0OSNative10.
100.
4.
0(OSNative)14.
0.
0.
7RevAHBAFirmware07.
02.
42.
003.
66(B)3.
2.
1.
54933N/AHBADrivers2.
0.
12.
206.
16.
0.
64(4Apr2011)3.
2.
1540.
60(8Sep2010)3.
1.
1548.
155Rev.
A22Oct2009ChipsetDriver9.
2.
0.
10.
21OSNative8.
73.
4(8Sep2010)1.
3.
0.
49VideoDriver1.
1.
3.
0OSNativeOSNative4.
1.
11.
1332IntegratedManagementControllerFirmware1.
70.
15BMC4.
22(A)7Apr2011NoiLOpresentNoiLOpresentManagementControllerDriverN/A1.
4.
0.
0N/AN/AMyFirstServer:TheDellPowerEdgeT110IIandHPEntryLevelServerComparisonAppendixD—Comparison1-4DetailedResultsForeachcomparison,thefirstpageoftheresultfilesforeachbenchmarkisshown.
SPECjbb2005andSPECpower_ssj2008benchmarkscontainresultsfileswhicharegeneratedwithgraphicandtabularresultsforeachserver.
Sandra2011SP2bonlyprovidesatextoutputfile,sothiswillnotbeincludedinthissection.
FullSandra2011SP2b,SPECjbb2005,andSPECpower_ssj2008resultsfilesforeachserverareattachedtothisdocumentforreference.
Figure6:Comparison1-SPECjbb2005ResultsforDellPowerEdgeT110IIMyFirstServer:TheDellPowerEdgeT110IIandHPEntryLevelServerComparisonFigure7:Comparison1-SPECjbb2005ResultsforHPProliantML110G6MyFirstServer:TheDellPowerEdgeT110IIandHPEntryLevelServerComparisonFigure8:Comparison1-SPECjbb2005ResultsforHPProliantMicroserverMyFirstServer:TheDellPowerEdgeT110IIandHPEntryLevelServerComparisonFigure9:Comparison1-SPECjbb2005ResultsforHPCompaq6005ProBusinessPCMyFirstServer:TheDellPowerEdgeT110IIandHPEntryLevelServerComparisonComparison2:Sandra2011SP2bCryptographicBandwidthresults.
TheresultsforSandra2011SP2bareoutputintotextformat,therearenographicalrepresentationsofthedata.
Theresultfileswillbeattachedtotheendofthisdocumentforreference,buttherearenovisualsthatcanbeshownforthissectionofthestudy.
Comparison3:SPECpower_ssj2008PerformanceperWattresults.
Theresultfilesforthesectionofthestudyareshownbelow.
Figure10:Comparison3-SPECpower_ssj2008ResultsforDellPowerEdgeT110IIMyFirstServer:TheDellPowerEdgeT110IIandHPEntryLevelServerComparisonFigure11:Comparison3-SPECpower_ssj2008ResultsforHPProliantML110G6Figure12:Comparison3-SPECpower_ssj2008ResultsforHPMicroserverMyFirstServer:TheDellPowerEdgeT110IIandHPEntryLevelServerComparisonFigure13:Comparison3-SPECpower_ssj2008ResultsforHPCompaq6005ProBusinessPCComparison4:Sandra2011SP2bStorageSubsystemBandwidthresults.
TheresultsforSandra2011SP2bareoutputintotextformat,therearenographicalrepresentationsofthedata.
Theresultfileswillbeattachedtotheendofthisdocumentforreference,buttherearenovisualsthatcanbeshownforthissectionofthestudy.