TRANSCODINGCHOICES FOR24x7x365LINEARVIDEO ATECHNOLOGYCOMPARISONBETWEENHARDWARE-BASED VIDEOCOMPRESSIONSYSTEMSANDSOFTWARE-BASED VIRTUALIZEDSERVERMODELS RICHARDPESKE VICEPRESIDENTOFVIDEOCOMPRESSION ANDPROCESSINGPRODUCTS
TABLEOFCONTENTS INTRODUCTION 3 THEUNIQUETRANSCODINGREQUIREMENTSOF 24x7x365LINEARVIDEO 4 ATECHNOLOGYCOMPARISONOFHARDWARE-AND SOFTWARE-BASEDTRANSCODINGFORLINEARVIDEO 5 Hardware-basedTranscoding 5 Software-basedTranscoding 6 SupportingtheSteady-stateWorkloadof Always-on Programming 7 IngestingVideofrom Various,OftenGeographicalyDistributed Sources 8 SupportingtheLong-term NeedsofLinearVideo 9 KeepingCapitalandOperationalCostsLow 10 CONCLUSION 12 REFERENCES 13 MEETOUREXPERT 13 Copyright2015 ARRISEnterprises,Inc.Alrightsreserved. 2
Withtherapidpaceofevolutionthat stakingplaceinthevideodeliverymarketplace,our industrymustconstantlyevaluatethebenefitsofnew technologiesandweightthem against thoseofcurrentmethods.thisespecialyholdstrueforvideoencodingandtranscoding, wheretraditionalhardware-basedmethodsarenow beingchalengedbyanew approachthat leveragespoolsofgeneral-purposeserversrunningsoftware.formanyvideoapplications suchasvod,this virtualized modelisgainingpopularityamongserviceprovidersasan efficientwaytoaccommodatetheconstantlychangingprocessingloadsofthoseapplications. However,thedynamicsoflinearbroadcastvideoaremuchdifferentfrom thoseofstored content,especialywhenitcomestotranscoding. Thispaperexaminestheuniquetranscodingrequirementsoflinearbroadcastvideoand compareshardwaresystemstonew virtualservermodelsinsupportingtheseneeds.init, ARRISexamineshow eachtranscodingchoicesupportstheprocessingworkload,channel volume,videoquality,andmultiscreendeliveryrequirementsoflinearvideo.inaddition,we examinethebusinessdynamicsofeachapproach,includingthechiefdriversofcapitaland operationalcostsforeachtranscodingmodel.atitsconclusion,thispaperprovidesaclear recommendationforserviceprovidersseekingguidanceontherighttechnologychoicefor linearvideotranscoding. INTRODUCTION Consumershavecometorelyonnetworkandpremium channelsasa go-to sourcefor entertainmentandinformation,whetherthey rewatchingonthelivingroom TVorusinga TVEverywhereapplicationononeofthemanydevicesattheirfingertips.Andwiththe evolutionfrom standarddefinition(sd)tohighdefinition(hd)to4kandultrahd(uhd), viewers expectationsforahigh-qualitylinearvideoexperienceareconstantlyincreasing.for serviceproviders,deliveringthisexperienceisatoppriority,butdoingsomustnotcomeat theexpenseofunreasonablecostsandmustnotexceedtheavailablecapacitywithintheir networks.thisiswheretherighttranscodingtechnologycanmakeorbreakthesubscriber experienceandtheserviceprovider sbusinesscase. Withtherighttranscodingmethods,serviceproviderscanpackthehighestqualityvideointo thefewestbits.thisisparticularlycriticalincaseswherethebandwidthavailableforvideo 3 WEB BLOG www.arris.com www.arriseverywhere.com
deliveryisascarceresource.examplesincludecableoperatorsthatarerunningoutofdocsis bandwidthorqam broadcastspectrum,telecommunicationsprovidersusingdsl,wireless operatorsrunning4goreven3gnetworks,anddirect-to-homesateliteproviderswithlimited capacity.inaddition,manyprogrammersarefacinglimitedtransponderbandwidthorseeking new waystoreducetheirvideotransmissioncosts. Ineachofthesecases,deliveringhighqualitylinearvideomustnotcomeattheexpenseof networkbandwidththatisn tavailableorisonlyatainableataprohibitivecost.thereisalso theneedtoconsideroperatingcostssuchaspower,cooling,andspace.operatorsareunder pressuretokeepcostsdownandtomeetnew green regulations,suchasenergy2020driven bythescteintheusortheenergyefficiencydirectivesfrom theeuropeancommissionfor 2020and2030.Andmostimportantly,thechosentranscodingapproachmustnotnegatively impactthesubscriberexperience.thekeytoaccomplishingaloftheseliesintheselectionof transcodingmethodsthataretailoredperfectlytofitthedistinctneedsoflinearvideo. Forthepurposesofthispaper,wewilusetheterm transcoding torefertoboththeprocess ofencodinguncompressedbasebandvideosignalsortranscodingpreviouslycompressed signals,unlessotherwisestated. THEUNIQUETRANSCODING REQUIREMENTSOF24x7x365 LINEARVIDEO Deliveringlinearvideoisanalways-onendeavor,andthatmeansasteadyworkloadforthe transcodersthatarepreparingeachstream fordistribution.muchlikethenews,shopping andsportsnetworkstheysupport,thesetranscodersareonline24hoursaday,sevendays aweek,365daysayear.thisdiffersfrom theroleoftranscodingforstoredvideoapplications suchasvideoondemand(vod).theseapplicationsplaceadynamicloadonthetranscoder, whichmustprocessvideoquicklywhen,forexample,anew movieortelevisionepisodeis addedtothevodlibrary.oncethetranscodeiscomplete,thereisalulinactivityuntilthe nextpieceofcontentisreadytobeprocessed.whenitcomestolinearvideodelivery,there isnorestperiod. Copyright2015 ARRISEnterprises,Inc.Alrightsreserved. 4
Inaddition,transcodersmustbeabletoadaptweltonew andupdatedcompressionstandards.asvideodistributionhastransitionedfrom SDtoHDtoUHD,compressionalgorithms haveevolvedtodeliverthisincreasingqualitywithoutover-taxingthenetwork.withthe migrationfrom AVCtoHEVCjustbeginning,transcodersmustnotonlysupportmultiple codecs,butadapttochangesinhow thenew algorithmsaredelivered.tosupportthelatest compressionstandards,transcodersmustbepowerfulenoughtohandleasignificantincrease inboththevolumeandcomplexityofcomputationsrequired. Theuniqueneedsoflinearvideocarrywiththem severalrequirementsforthetranscoding methodsusedtoprocessit.tosupporttheconstantneedsof24x7x365videodeliverywhile maintainingahighqualityuserexperience,today stranscodersmustperform ataconsistently highlevel,whilebeingscalable,efficientandadaptabletochange.inaddition,therighttranscodingapproachforlinearprogrammingmustalsomakeeconomicsensewhenitcomesto capitalandoperationalexpenses.eachoftheserequirementswilbeusedtocomparehardware- basedsystemsandsoftware-basedvirtualizedserversfor24x7x365linearvideotranscoding. ATECHNOLOGYCOMPARISON OFHARDWARE-ANDSOFTWARE- BASEDTRANSCODINGFOR LINEARVIDEO Tocomparethetwomostcommonapproachesforlinearvideotranscoding,itisimportant tofirstdefinethem. Hardware-basedTranscoding Broadcastqualityvideocompressionhastraditionalybeenperformedwithhardwaresystems builtwithspecializedcompressionsilicon.thespecializedcompressionsiliconchipsusedin thesesystemsutilizededicatedcircuitstoperform thecomplexandspecializedcomputations ofvideoprocessing.thesesystemsarebuiltwithavarietyofspecializedinterfacesthatcan ingestvideoinarangeofformats,withthelatesttranscoderarchitecturesutilizingamodular designthatalowstheseinterfacestobechangedasneeded. 5 WEB BLOG www.arris.com www.arriseverywhere.com
HARDWARE-BASEDTRANSCODING SOFTWARE-BASEDTRANSCODING Figure1:ASimplifiedViewofHardware-basedandSoftware-basedTranscodingSystems Software-basedTranscoding Software-basedtranscodingleveragestheever-increasingpowerofgeneralpurposecomputingserverstoprocessvideo.Ageneralpurposeserverfeaturesal-IPinterfacesandhasa powerfulx86multi-coresiliconchipthatisdesignedtohandleawiderangeofcomputing tasks.multipleserverscanbeconnectedintoavirtualizedprocessingpool,whichcanbe alocateddynamicalytotheencodingtaskusingsoftwareasneeded.server-basedtranscoding hasseenanincreaseinadoptionbasedonitsuseofoff-the-shelfserversanditsflexibilityto supportrapidlychangingworkloads. Thereisalotofdiscussionaboutusinghardwaretoassistgeneralpurposeservers.Thiscan beintheform ofaddedprocessingcardsusingeitherdedicatedsiliconorfieldprogrammable GateArays(FPGA).Butthesetypesofdevicesshouldbeclassifiedashardware-basedsystems asfarasvideocompressionisconcerned.alternatively,intelisincorporatinggraphicalprocessingunits(gpu)andsomefixedfunctionhardwarefortaskssuchasvideocompression intoitscpuprocessorline.however,serversthatutilizethesespecializedprocessorsarenot generalyfoundinstandarditdatacenters,althoughtheiruseisexpectedtoincrease.in addition,thesehardwareblocksaretypicalyfocusedonlowerqualityinternet-stylevideo,so theirusewithprocessingofbroadcastqualityvideoforhighresolution,largescreendisplay devicesislimited.inthispaper,wewilfocusongeneralpurposedatacenterserversusedfor virtualizedcloudapplications,unlessspecificalynotedotherwise. Copyright2015 ARRISEnterprises,Inc.Alrightsreserved. 6
Byusingtheuniquelinearvideotranscodingrequirementsasaguide,wecannextanalyze how software-basedmodelscomparetohardwaresystemsinprocessingvideocontentfor 24x7x365delivery. SupportingtheSteady-stateWorkloadof Always-on Programming Theconstantcompressionrequirementsoflivelinearvideoarequitedifferentfrom the dynamicworkloadstypicalofthetranscodingofstoredcontentorevenoccasionalliveevents. Thisneedforcontinuousprocessingeliminatesoneoftheprimarybenefitsofvirtualized transcoding:theefficiencygainedfrom dynamicalyalocatingcomputingpoweracross multiplevirtualmachinesasprocessingneedsfluctuate.conceptualy,onecouldadjustthe amountofprocessingpowerinproportiontothecomplexityofthevideo,forexampleusing morecpuresourcesforcomplexvideosequences,buttheoverheadoftheresourcemanagementimposedbyvirtualmachinesoftwareistoohighforreal-timevideoprocessing.itis muchsimplertoalocateafixedamountofprocessingresourcetoalivelinearstream,thus negatingamajoradvantageofvirtualization. 60 50 LinearVideo VoD 40 30 20 10 0 8:00AM 10:00AM 12:00PM 2:00PM 4:00PM 6:00PM 8:00PM 9:00AM 11:00AM 1:00PM 3:00PM 5:00PM 7:00PM Figure2: Comparingthe Transcoding Workloadsfor VoDandLinear Video Conversely,hardware-basedtranscoding,withitsdedicatedprocessingarchitecture,ismuch betersuitedtothesteady-stateneedsoflinearvideo.bymaintainingaone-to-onerelationship betweenalinearchannelanditstranscodingresource,theappropriateamountofcomputing 7 WEB BLOG www.arris.com www.arriseverywhere.com
powercanbeappliedtoprocessvideoforagivenchannel.theabilitytodedicateresources notonlytovideoencodingbuttootherprocessingtaskssuchasvideodecoding,audioand metadata,makeshardware-basedcompressionsystemsmuchmorepredictableandmakes resourceplanningmuchsimpler.thispredictabilityisevenmoreimportantwiththeaddition ofthemultipleprofilesrequiredformultiscreenvideodelivery.itcanbeveryhardtopredict theimpactofprocessinghigherbit-ratemezzanineinputsortheadditionofanew video profileforadaptivebit-rate(abr)environments. HARDWARE-BASEDTRANSCODING SOFTWARE-BASEDTRANSCODING VS ADVANTAGE:HARDWARE IngestingVideofrom Various,OftenGeographicalyDistributedSources Therearetwoprimaryconnectivitychalengespresentedbylinearvideo:colectingvideo streamsefficientlyfrom localprogrammersorbroadcastersthatareoftenalongdistance away,andensuringthattheycanbeingestedeasilyoncetheyreachthetranscoder.in evaluatinghardwareandsoftwaretranscodingapproaches,thesetwofactorsarestrongly impactedbytheinterfacesusedtoacquirevideosignalsfrom thenetwork. GenericserversforvirtualizedcloudapplicationsuseIPinterfacesexclusivelytoingestvideo. ButdeliveringuncompressedHDandUHDvideostreamscanrequiresignificantamountsof expensiveipnetworkbandwidth.forhdvideo,eachstream requires1.485gbpsat1080i30, andtwicethatfor1080p60.theserequirementsincreaseto11.88gbpsforanuncompressed 4Kp60stream.TodeliverhighqualityvideooveranIPnetworkcosteffectively,thesevideo streamsmustbepre-compressed.toenablethismodel,serviceprovidersmustdeployand manageaseriesofgeographicalydistributedanddedicatedvideocompressionresources, whicharenotsuitableforvirtualizationincentralizedorregionaldatacenters. Linearstreamsaretypicalyencodedandtransportedfrom theprogrammer/broadcasterto theserviceproviderinahigh-bitratemezzanineformatinordertomaintainthehighest possiblevideoquality.consequently,software-basedsystemsneedtodedicatemoreprocessingresourcestodecodethesetypesofstreams,whichwilreducetheresourcesavailable Copyright2015 ARRISEnterprises,Inc.Alrightsreserved. 8
Iforencoding.Newerhardware-basedsystemsarenottypicalyimpactedbythetypeofdecodingrequired,sincetheencodinganddecodingfunctionsuseseparateresourcepools. Geographicalydistributedcontentandmultipleingestformatsarewelhandledbycompressionapplianceswithspecializedtranscodinghardwareplatforms,whichcomeequippedwith amultitudeofinputstoingestvideoinanyuncompressed,mezzanineorhighlycompressed format.withahardware-basedtranscoder,serviceproviderscancustomizetheirtranscoding architecturetoingestvideointheformatsthatworkbestfortheirprogrammingpartnersand distributionnetworks. HARDWARE-BASEDTRANSCODING SOFTWARE-BASEDTRANSCODING VS ADVANTAGE:HARDWARE SupportingtheLong-term NeedsofLinearVideo Intheworldofvideodelivery,onethingiscertain:linearvideoisn tdisappearing.livebroadcastswilbearoundaslongastherearetelevisedsportingevents,24-hournewsnetworks andpay-per-view specials.butevenasasteadyforce,thedeliveryoflinearvideoisundergoing change.new compressionalgorithmsareenteringthefoldwithincreasingfrequency,and existingstandardsareevolvingtobemoreefficientandstable.whenchoosingbetweenhard- ware-andsoftware-basedtranscodingapproaches,wemustconsiderthelong-term needsof linearvideo. Thelatestgenerationhardware-basedtranscodersaredesignedandbuilttolastforwelover adecade.fulymodular,thesesystemscanevolvewithouttheneedforaforkliftupgradeor re-cablingintheequipmentrack.thiscommitmenttolongevityisupheldbysiliconsuppliers andvendorengineeringteams,whounderstandthatvideotranscodingsystemsneedto remainup-to-dateandinproductionformanyyears.thisextendedlifecycleisafundamental differentiatorbetweenitdataprocessingequipmentandbroadcastvideoequipment. 9 WEB BLOG www.arris.com www.arriseverywhere.com
GenericServers7years Figure3:Comparingthe LifespansofGeneric ServersandHardware Transcoders Lifespan inyears HardwareTranscoders10years 0 2 4 6 8 10 12 HARDWARE-BASEDTRANSCODING SOFTWARE-BASEDTRANSCODING VS ADVANTAGE:HARDWARE KeepingCapitalandOperationalCostsLow Inconsideringtheuniquetranscodingneedsofdeliveringlinearvideo,wecannotexclude thepracticalrequirementforaneconomicalyviablesolution.andwhile,onthesurface, genericserverscanbepurchasedmoreinexpensivelythanspecializedtranscodinghardwareplatforms,wemustinvestigatethetotalcostsofownershipforthetwomodels. Thiscalsforananalysisofthedriversforboththecapitalequipmentcostsaswelasthe ongoingoperationalexpensesassociatedwitheachtranscodingapproach. Whilegenericserversalonecanbepurchasedmoreinexpensivelythanhardware-based transcoders,bothhardwareandsoftwareexpensesmustbefactoredinforwhencomparingthecapitalcostsperchannelofthetotalsolution.theseper-channelsolutioncostsare drivenprimarilybycompetitiveforces,andthereforeareverysimilarforhardware-and software-basedarchitectures.however,whenexaminingpowerandrealestateutilization forbothapproaches,theoperationalexpensesofeachmodelarequitedistinct. Copyright2015 ARRISEnterprises,Inc.Alrightsreserved. 10
Tocomparetheoperationalcostsofhardware-andsoftware-basedtranscoding,weutilizea 5-year(43,800-hour)timehorizon,assumeacostperKilowat-Hourof$.10andassignapower utilizationefficiency(pue)of2.5forbothmethods.thismetricpresumesaslightimprovement inpuefordatacenteroperationsovertheblendedindustryaverageof2.91.inthiscomparison, wealsoassumeachannelcountof3,500todemonstratethecompoundedoperationalcosts foraserviceproviderwithabroadgeographicalreachandtheneedtotranscodemultiple videoprofilesformultipledisplaytypes.weassumetheuseofavc/mpeg-4transcodingas thisisthemostcommonformatforip-baseddelivery. GenericServer 2015TECHNOLOGY HardwareTranscoder PowerDissipationperchannel 44 8.3 Costforonechannel $481.80 $90.89 Costfor3,500Channelsover5Years $1,686,300.00 $318,097.50 TotalCostAvoidanceofHardware TranscodersoverGenericServers $1,368,202.50saved Table1:ComparingOperationalCostsforGenericServersandHardwareTranscoders AsTable1shows,thepowerexpensesavingsofhardwaretranscodingaresubstantialwhen comparedtogenericserverimplementations,reducingpowerexpendituresbyover80%,a majordifferenceinenergyefficiency,whichisnow amandateformostoperators. Inadditiontothepowercostsoftranscoding,realestateutilizationisalsoacriticalfactorintoday s space-limitedvideodeliveryfacilities.table2comparestherackunitrequirementsfortrans- codingthesame3,500linearchannelsusinggenericserversandspecializedhardware.once again,hardwaresystemsprovideasignificantadvantage,achievingaspacesavingsofover70%. 2015TECHNOLOGY GenericServer HardwareTranscoder ChannelsperRackUnit 12 48 RackUnitsRequiredfor3,500Channels 294 77 SpaceSavingsofHardware TranscodersoverGenericServers 217RackUnitssaved Table2:ComparingRealEstateUtilizationforGenericServersandHardwareTranscoders 11 WEB BLOG www.arris.com www.arriseverywhere.com
Theefficienciesofhardwaretranscodersareclear.Inadditiontoreducedcapitalcosts, hardwareimplementationsconsumemuchlesspowerandtakeuplessfloorspacethan genericservers.whilenotdepictedinthispaper,additionalopportunitiesforsavingsusing hardware-basedtranscodingincludereducedcoolingexpenses,fewermanagementstaff required,andthelow maintenancecostsassociatedwiththe99.999% reliabilityofhardware transcoders.andlastly,astheindustrybeginsmovingtonewer,significantlymorecomputa- tionalycomplexcompressionalgorithmssuchashevc(typicaly5xto10xtheencoding complexityofavc),theadvantagesofhardware-basedsystemsareevenmoremagnified. HARDWARE-BASEDTRANSCODING SOFTWARE-BASEDTRANSCODING VS ADVANTAGE:HARDWARE CONCLUSION Whenitcomestotheuniqueneedsof24x7x365linearvideo,hardwaretranscodersprovide aclearadvantageovervirtualizedsoftwaretranscoders.hardwaretranscodersprovidethe dedicatedprocessingneededtosupportthesteady-stateworkloadsof always-on programming.theyareequippedwiththeinputflexibilityneededtoingestvideofrom multipletypes ofsources,whileusingnetworkbandwidthefficiently.hardwaretranscodersalsoenablethe flexibilityandlongevityneededtokeeppacewithlinearvideo sever-evolvingneeds.andthey doitcost-effectively,withlowercapitalandoperationalexpensesderivedfrom greaterdensity, beterpowerutilizationandadditionalsavingsassociatedwithgreateruptimeanddecreased coolingrequirementswhencomparedtothecostsofsoftware-basedtranscoding. Copyright2015 ARRISEnterprises,Inc.Alrightsreserved. 12
REFERENCES 1 Niccolai,James. New datacentersurveyshowsmediocreresultsforenergyefficiency. Computerworld,12April2013.Web24Mar.2014. MEETOUREXPERT:RICHARDPESKE AsVicePresidentofVideoCompressionandProcessingProducts atarris,richardleadstheproductmanagementteam forthe company smultiscreenandnext-generationhevcsolutions.in thisrole,hedrivesthedefinitionanddevelopmentofadvanced hardwareandsoftwareproductsthatmaximizevideoqualityfor endusers,whileminimizingthecosts,complexitiesandnetwork impactsofvideodeliveryforserviceproviders.richard s20+years ofexperienceindigitalvideodeliveryincludesindustrypioneers suchasbigbandnetworks,divicom,silicongraphicsandbellabs. HeholdsanMBAfrom thekeloggschoolofmanagementat NorthwesternUniversity,anM.SdegreeinElectricalEngineering from theilinoisinstituteoftechnologyandab.sdegreein ElectricalEngineeringfrom theuniversityofmichigan ARRISEnterprises,Inc.2015Alrightsreserved.Nopartofthispublicationmaybereproducedinanyform orbyany meansorusedtomakeanyderivativework(suchastranslation,transformation,oradaptation)withoutwriten permissionfrom ARRISEnterprises,Inc.( ARRIS ).ARRISreservestherighttorevisethispublicationandtomake changesincontentfrom timetotimewithoutobligationonthepartofarristoprovidenotificationofsuchrevision orchange. 13 WEB BLOG www.arris.com www.arriseverywhere.com