productsxendesktop

LabTestingSummaryReportKeyfindingsandconclusions:InacomparisonofVirtualDesktopInfrastructure(VDI)implementations,CitrixXenDesktop4providedbetteroverallperformancewhencomparedtoVMwareView4XenDesktop4used64%lessbandwidththanView4withPCoIPfortypicaltasksFlashvideowasdeliveredwithanaverageof65%lessCPUusage,89%lessbandwidth,andexcellentQualityofExperiencebyXenDesktop4comparedtoView4Overall,XenDesktop4usessystemresourcesmoreefficientlyandiscapableofscalingmoreeffectivelyFebruary2010Report100207ProductCategory:VirtualDesktopInfrastructureVendorsTested:ProductsTested:CitrixXenDesktop4VMwareView4CSinceXenDesktop4EnterpriseandPlatinumeditionsoffermanyotherformsofdesktopandapplicationvirtualizationnotavailableinView4,thesetestsfocusedprimarilyonVDIinordertoenableafaircomparison.
Specificallyevaluatedwerethesystemefficiency(bandwidthandCPUconsumption)andQualityofExperience(QoE)fortwotypesofworkloads:[1]real-world,everydaydesktoptasks,and[2]popularformsofmultimediaandrichInternetapplicationssuchasFlash.
TheseworkloadswereevaluatedunderbothLANandWANconditionstoassesstheimpactofbandwidthandlatency.
WeverifiedtherealworldapplicationresponsivenessforthevirtualclientsusingsimulationwithautomatedscriptsthatincludedMSWord,MSExcelandMSOutlook.
itrixSystemsengagedMiercomtovalidatetheoveralleffectivenessofsimilarlyconfiguredVirtualDesktopInfrastructure(VDI)solutionsusingCitrixXenDesktop4andVMwareView4.
OurtestresultsshowedthatCitrixXenDesktop4usedupto64%lessbandwidthcomparedtoVMwareView4whileaccessingrealworldapplicationslikeMSWord,Excel,InternetExplorerandOutlook.
Source:Miercom,January2010Figure1:BandwidthUtilizationforReal-TimeDesktopWorkload0.
3771.
0290.
000.
200.
400.
600.
801.
001.
20CitrixXenDesktop4VMView4Bandwidth(Mbps)CitrixXenDesktop4VMwareView4Copyright2010MiercomVirtualDesktopInfrastructurePage2BandwidthConsumption:GeneralDesktopWorkloadThistestwasconductedtodemonstratethebandwidthconsumptionwhenarealworldworkload(MicrosoftOutlook,Excel,Word,PowerPoint;Flash-basedwebsitesandmultimedia;PDFprintingandviewing)isaccessedusingavirtualdesktopplatform.
WesimulatedtheWANandLANenvironmentsasexplainedinthetestbedtogeneratethebandwidthconstraintstomatchrealworldscenarios.
Themetricsofthetestshowtheaveragebandwidthconsumptionfortheapplicationsrunningonvirtualdesktopplatformsbasedondifferentdeliveryprotocols.
Inthefirstpartofthetest,weexecutedLoginConsultantsVSI2.
0scripttosetthebenchmarkforVirtualDesktopInfrastructureservicesonbothoftheservers.
WeusedthisautomationframeworkforimplementingtheworkloadontheserverwhenusingtheremoteclienttoaccesstheseservicesthroughVDIinterfaces.
Duringthistest,wemeasuredthebandwidthconsumptionforbothXenDesktop4andVMwareView4asshowninFigure1onpage1.
Thisisthetransmissionrateatwhichtheclientissendingandreceivingdataoverthenetworktotheserver.
WeusedClearSightNetworkAnalyzertoassessprotocolefficiencyinclient/servercommunications.
FindingsOnanalyzingtheprotocoltraces,wefoundthatXenDesktopusedTCP,whileView4usedprimarilyUDP,aswellasTCP.
OnanalyzingtheClearSighttracesforprotocolhierarchy,wefoundthattheVMware'sPCoIPprotocolwasconstantlysendingandreceivingpacketsattherateof1.
093MbpswhileXenDesktop,whichusestheCitrixICAprotocol,wastransmittingpacketsattherateof0.
377Mbpsbetweenclientandserver,resultinginlessbandwidthconsumptionforXenDesktop4.
WeobservedconsiderablymoreoverheadtrafficintheVMwareView4environmentwhencomparedtoCitrixXenDesktop4.
DuringtestingweobservedpeakmeasurementsfortheVMwareenvironmentwheretheaddedoverheadbetweenserverandclientoverloadednetworkresourcesandimpairedtheapplicationqualityrunningonthevirtualclient.
AsseeninFigure3onpage3,thecommunicationbetweentheclientandserverforCitrixXenDesktop4consumedanaverageof377Kbpsofbandwidth,whileVMwareView4consumedanaverageof1.
093Mbpsofbandwidthforthesametestenvironment.
UponanalyzingthetrafficcapturesforbothVDIproducts,XenDesktop4had66,945,228Bytes,whileVMwareView4hadfourtimesmoretrafficat274,726,011Bytes(fullEthernetFrames,headerandpayload).
ImpactForanyorganizationlookingtoimplementVDIatenterprisescale,thesetestresultsindicatethatXenDesktopwillenableseveraltimesmoreusersthanView4foragivennetworkcapacity.
ThedesktopworkloadgeneratedbyLoginVSIprovidesagoodrepresentativesamplingforhowrealuserswillstressthesystem.
Thetestsshowthatasusersswitchfromcommonproductivityandline-of-businessapplications,tomoredemandingwebbrowsingwithFlashmedia,CitrixXenDesktopwillusenetworkresourcesmuchmoreefficiently,whilestillofferinghighqualityuserexperience.
Thiswilltranslateintoloweroverallcosts.
QualityofExperience:GeneralDesktopWorkloadDuringtherunningoftheLoginVSItestscripts,wealsovideo-capturedandobservedthequalityandresponsivenessofthescreendisplay.
WhileXenDesktopmaintainedaverylocal-likeappearancewhenrunningapplicationslikeMSExcel,Viewdisplayedsomeevidentblurringthatmadethetextdifficulttoread.
Weattributethistothe"buildtolossless"featureofPCoIP.
Whilethisfeaturemayaidinsomegraphicapplicationsandmedia,usersmayfind,justasweobservedinourtesting,thatthisisdistractingforsomeapplications.
Forasampletestvideooftheuserexperience,emailyourrequestto:reviews@miercom.
comFigure2:BandwidthUtilizationwhenStreamingFlashVideoTestresultsshowedthatCitrixXenDesktop4consumedonly3.
008Mbpsofbandwidthcomparedto29.
04MbpsofbandwidthconsumptionbyVMwareView4.
293051015202530CitrixXenDesktop4VMView4Bandwidth(Mbps)CitrixXenDesktop4VMwareView4Source:Miercom,January2010Copyright2010MiercomVirtualDesktopInfrastructurePage3BandwidthConsumption:FlashVideoThistestwasconductedtodemonstratethebandwidthconsumptionwhileplayingFlashvideosavailableonInternetwebsiteslikewww.
hulu.
comorwww.
youtube.
com.
SincebothXenDesktop4andView4supportFlashstreamingonvirtualdesktops,andhavecapabilitiestooptimizemultimedia,wetestedthebandwidthconsumptionaswellastheQualityofExperience.
Forthistest,westreamedFlashvideosfromwebsitesincludingwww.
hulu.
comandwww.
youtube.
comwhenloggedinasaremoteuseronathinclienttobothXenDesktop4andView4servers.
Asinprevioustests,werepeatedmeasurementsinbothamplebandwidthaswellasconstrictedbandwidthenvironments.
AnAppositeWANemulatorwasusedtosimulatetheimpairednetworkconditionsthatcouldoccuroverenterpriselevelLAN,WANandremoteaccessenvironments.
WeintroducedWANlatencyinrangesof100-250msandpacketlossof0.
5-5%onthe100MbpsEthernetlinks.
FindingsFigure2onpage2illustratesthetestresultsinwhichCitrixXenDesktop4consumedanaveragetransmissionrateof3Mbps,whileVMwareView4consumedamuchgreateraveragetransmissionrateof29MbpsoveraperiodofthreeminutesFlashvideostreaming.
Wecross-checkedthedatafromtheWANemulatorwiththatoftheClearSightNetworkAnalyzerreportfortheaveragetransmissionrateovertheWAN.
AsshowninFigure4onpage4,VMwareView4hasacontinuoushighvolumeoftrafficbecausetheFlashvideoisrenderedontheserverandthenpushedoverthenetwork.
contrast,CitrixXenDesktop4sendsthecompressed,non-renderedvideototheendpointforrenderinglocally.
QualityofExperience:FlashVideoInadditiontobandwidthconsumption,wealsoassessedvideoqualityforeachproductwiththeirrespectivemultimediastreamingoptimizationsasexplainedintheprevioussection.
Avideomeanopinionscore(VidMOS)ratingfortheQualityofExperience(QoE)wasdeterminedforeachproductundertest.
ThetabledescribingQoEandvideoqualityratedbyvideoMOSrelativetoobservablevideoimpairmentsisshowninFigure5.
Westreamedvideosfromwebsitesincludingwww.
hulu.
comandwww.
youtube.
com.
FlashvideostreamsavailableontheWorldWideWebwereplayedconsecutivelyfor5minuteswithnoWANimpairmentsintroduced.
WescoredeachvideostreamedonthescaleasdescribedinFigure5onpage4,usingVidMOScriteriaforscoring.
WithFlashoptimizationdisabled,CitrixXenDesktop4scoredVidMOSof4.
0whereasVMwareView4exhibited"burstier"andlessconsistentvideoqualitywithmanydistortions,andonlyachievedaVidMOSscoreof2.
0.
AverageTxrateshowedthatinatimespanof10minutes,377KbpsofbandwidthwasconsumedbyCitrixXenDesktop4,whereasVMwareView4consumed1.
029Mbps.
Figure3:VirtualDesktopInfrastructureBandwidthConsumptionforTypicalOfficeApplicationsSource:Miercom,January2010CitrixXenDesktopBandwidthConsumptionVMwareViewBandwidthConsumptionCopyright2010MiercomVirtualDesktopInfrastructurePage4ThisstreamingwasdonewiththeavailableFlashoptimizationfeatureenabledforbothCitrixXenDesktop4andVMwareView4.
WefoundthatwhiletheFlashoptimizationsettingforVMwareView4improvesthevideoquality,itsignificantlyincreasesbandwidthutilizationasshown.
Thevideoqualitytestwasrepeated,thistimewithFlashAccelerationenabledforXenDesktop4,AdobeFlashQualitysettohigh,andAdobeFlashThrottlingsettoConservativeforVMwareView4PCoIP.
ThesesettingsallowedforthebestvideoexperiencewithVMwareView4;howeverwefounditwasstillinferiortotheQoEachievedwithCitrixXenDesktop4.
WeobservedimprovementinvideoqualityforVMwareView4withPCoIPbutitwasattheexpenseofgreaterbandwidthandserverCPUutilization.
WescoredthevideoqualityforVMwareView4aVidMOSscoreof3.
0forthevoiceandvideostreamed.
Wenotedalackofsynchronizationbetweenthevideoandaudio,andsomenoticeablevideodegradation.
VidMOSscoreswiththetestandmeasurementequipmentachievedbetween2.
9and3.
1.
ForCitrixXenDesktop4thestreamingFlashvideoachievedaperfectscoreof5.
0ratedbothbythetestandmeasurementtoolsaswellasouruserobservablemeanopinionscores.
Ascoreof5.
0representsthatthedeliveredvideowasidenticaltothesourcevideostream,andtherewerenoobservableimperfections.
AdditionaltestswereconductedforVMwareView4toallowforothersettingssuchaslowtohighforAdobeFlashQualityandaggressivetoconservativeforAdobeFlashThrottling,buttherewasnonoticeablechangeinvideoqualityforVMwareView4.
WethenusedtheClearViewQAtoolforthequantitativeanalysisofthevideoscapturedduringthesetests.
Theaveragescoreof4.
8forXenDesktop4and2.
9forView4wasnotedasshowninFigure6onpage5.
Constrainingthebandwidthto10MbpswiththeAppositeWANemulatorwithalatencyof250msecandpacketlossof1percent,weobservedthatthequalityofXenDesktop4wasstillnotaffected;achievinga4.
0VidMOSscore,whilethequalityofView4significantlydroppedtoascoreof1.
1VidMOS(unusable).
Tobetterunderstandthequalityofa1.
1VidMOSvideo,emailyourrequestforaFlashvideosampleto:reviews@miercom.
comCPUUtilization:FlashVideoCPUutilizationwasmeasuredforbothCitrixXenDesktop4andVMwareView4whilestreamingHD-qualityvideos.
WemeasuredtheCPUutilizationofthevirtualmachinesinstalledontheserverforbothXenDesktop4andView4.
WeplayedthesameFlashvideofor10minutesatthe100Mbpslinerate.
WenoticedthatCitrixScoreQualityImpairments5ExcellentNotNoticeable4Good3FairSomeNoticeable2Poor1BadAnnoyingandUnusableFigure5:MeanOpinionScore-QoEforVideoQuality(VidMOS)VideoMeanOpinionScore(VidMOS)tableforQualityofExperience(QoE)andvideoqualityratedbyvideoMOSrelativetoobservablevideoimpairments.
Source:Miercom,January2010Figure4:VirtualDesktopInfrastructureBandwidthConsumptionforFlashVideoStreamingCitrixXenDesktopBandwidthConsumptionforFlashVideoStreamingVMwareViewBandwidthConsumptionforFlashVideoStreamingCopyright2010MiercomVirtualDesktopInfrastructurePage5XenDesktop4consumedanaverageof4.
7%processortime,whileVMwareView4consumedanaverageof71.
5%processortime.
AsshowninFigure7onpage5,wenoticedthatseveraltimesduringthetestingofView4,theprocessorutilizationspikedto100%,whichaffectedthevideoqualitysignificantly.
CitrixXenDesktop4hasauniquefeaturecalledHDXMediaStreamthatleveragestheprocessingpoweroftheendpointdevice.
Thisfeaturerendersthemultimediacontent;therebyreducingtheburdenontheserverCPUprocessor.
Thecompressedmultimediainformationissentdirectlytotheendpointinitsnativeformat.
Themultimediastreamisrenderedandplayedbacklocally,providingexcellentperformancewhilereducingtheworkloadontheserversandthenetwork.
ImpactXenDesktop,withHDXfeaturesforoptimizingFlash,allowsformuchmoreefficientuseofresourcesallocation,includingupto90%lessbandwidthand65%lessCPUutilization.
EvenwithViewFlashoptimizationsettingsenabledandadjustedtoseveralpossibleconfigurations,anyimprovementsinexperiencestillresultedinmorebandwidthandCPUconsumption.
Withlessresourcesrequiredpervirtualdesktop,XenDesktop4allowsforgreaterscalabilityperserver.
BottomLineOurtestsclearlyprovedthatCitrixXenDesktop4consumedlessbandwidthandCPUresourcesfordeliveringcommondesktopworkloadsandhighqualityFlashvideoswhencomparedtoVMwareView4.
TheresponsivenessofapplicationsonthevirtualmachineswasnotablyfasterandimproveduserQoEwithXenDesktopthanVMwareView.
Inaddition,throughoutourtestingwenotedthatdeploymentandpolicymanagementusingCitrixXenDesktop4provedeffectiveandsuitableforlargeenterprises.
CPUutilizationmeasuredwhiletheserversprovidingthevirtualdesktopenvironmentswereservingvideostreamingcontenttothevirtualdesktops.
FlashoptimizationfeatureswereenabledforbothVMwareView4andCitrixXenDesktop4.
CitrixXenDesktop4achievedaVidMOSscoreof4.
8withstreamingFlashvideowhereasVMwareView4scored2.
9.
ThesearequantitativescoringofFlashvideosqualityprovidedbyVideoClarityClearViewtool.
Figure6:VideoMOSscoreswithStreamingFlashVideo2.
94.
80123456CitrixXenDesktop4VMwareView4VidMOSScoreSource:Miercom,January2010CitrixVMwareXenDesktop4View4Figure7:CitrixXenDesktopBandwidthConsumptionComparisonSource:Miercom,January2010CitrixXenDesktop4VMwareView4%CPUUtilizationCopyright2010MiercomVirtualDesktopInfrastructurePage6TestBedDiagramTestEquipmentUtilizedinTesting:WeusedtwoHPDL360serverswith2IntelXeonQuadCoreprocessorsforthetesting.
Eachserverhad12GBofRAMand3x72GBdiskdrivesforinstallingtheXenDesktop4andVMwareView4suiteforVDItesting.
TheserverwithXenDesktop4wasloadedwithXenServerversion5.
3.
3.
1,ServiceConsole3.
0,DesktopDeliveryControllerVersion4.
0,VirtualDesktopAgent4.
0.
4522,VirtualDesktopClient11.
2.
0.
31560.
ThesecondserverwasloadedwithacopyofVMware4.
0.
0,ViewConnectionServer4.
0.
0-210399,ViewAgent4.
0.
0-210939,ViewClientwithoffline4.
0.
0.
-210939.
HowWeDidItAsshowninthetestbeddiagram,weconnectedtwoserverstotheAppositeWANemulatorthroughaswitch.
WeusedanAppositeWANemulatorwww.
apposite-tech.
comforbandwidthrestrictionandtointroducepacketlossandlatencytothenetworktosimulateremoteuserconnectionsandconstrictedWANenvironments.
WealsousedLoginConsultantsVSIscriptwww.
loginconsultants.
comthathelpedautomatethelaunchofMicrosoftapplicationsincludingExcel,WordandOutlookinaconsistentmannerwhilemeasuringCPUandbandwidthutilization.
WeusedtheClearSightNetworkAnalyzerwww.
clearsightnetworks.
cominourtestbedtoanalyzetheefficiencyoftheapplicationprotocolsused:ICAwithHDXforCitrixandPCoIPforVMware.
WeutilizedVideoClarityClearViewsystemswww.
videoclarity.
comtodetermineVidMOSscoresforquantitativeanalysis.
Eachserverwasequippedwithadomaincontroller,DNS,connectionserverandbrokerasthepartofthevirtualizationplatform.
WemeasuredbandwidthutilizationandassessedtheQualityofExperience(QoE)whileplayingFlashmoviesaccessedthroughInternetExplorer8fromwebsitesincludingwww.
hulu.
comandwww.
youtube.
com,usingthevirtualdesktopsoftwareinstalledoneachclient.
WeusedanIxiaXM2chassiswww.
ixiacom.
comtoapplyIMIXbackgroundtrafficwhilecapturingtheFlashvideocontentforanalysis.
Real-worldtrafficwasalsousedintestingasgeneratedbyIxia'stestplatformandtestapplications,principallyIxNetworkforLayer2-3routingandswitchingtrafficandIxLoadforLayer4-7applicationtraffic.
SwitchVMwareView4ServerLANALANBAppositeLinktropy7500PROVirtualClientClearSightNetworkAnalyzerCITRIXXenDesktop4ServerIxiaXM2ClearViewQACopyright2010MiercomVirtualDesktopInfrastructurePage7Productnamesorservicesmentionedinthisreportareregisteredtrademarksoftheirrespectiveowners.
Miercommakeseveryefforttoensurethatinformationcontainedwithinourreportsisaccurateandcomplete,butisnotliableforanyerrors,inaccuraciesoromissions.
Miercomisnotliablefordamagesarisingoutoforrelatedtotheinformationcontainedwithinthisreport.
ConsultwithprofessionalservicessuchasMiercomConsultingforspecificcustomerneedsanalysis.
AboutMiercom'sProductTestingServicesReport100107reviews@miercom.
comwww.
miercom.
comMiercomhashundredsofproduct-comparisonanalysespublishedovertheyearsinleadingnetworktradeperiodicalsincludingNetworkWorld,BusinessCommunicationsReview-NoJitter,CommunicationsNews,xchange,InternetTelephonyandotherleadingpublications.
Miercom'sreputationastheleading,independentproducttestcenterisunquestioned.
Miercom'sprivatetestservicesincludecompetitiveproductanalyses,aswellasindividualproductevaluations.
Miercomfeaturescomprehensivecertificationandtestprogramsincluding:CertifiedInteroperable,CertifiedReliable,CertifiedSecureandCertifiedGreen.
ProductsmayalsobeevaluatedundertheNetWORKSAsAdvertisedprogram,theindustry'smostthoroughandtrustedassessmentforproductusabilityandperformance.
Beforeprinting,pleaseconsiderelectronicdistributionCitrixSystems,Inc.
851WestCypressCreekRoadFortLauderdale,FL33309(954)267-3000www.
citrix.
comCitrixXenDesktopVersion4MiercomPerformanceVerifiedBasedonMiercom'sreviewofVirtualDesktopInfrastructure(VDI)applications,theCitrixXenDesktop4isawardedPerformanceVerifiedforprovidingbetterperformancewhencomparedtoVMwareView4.
XenDesktopconsumed64%lessbandwidththanViewwithPCoIPduringtestsconductedusingtypicalofficeapplications.
Flashvideowasdeliveredwithanaverageof65%lessCPUusage,89%lessbandwidth,andbetterQoEbyXenDesktop.
CitrixXenDesktop4enablesbusinessestodeployVDIinlargerscaleandsupportsFlashvideofarbetterthanVMwareView4.