mobicom11-peng-greenbs-final.ppt

Traffic DrivenPowerSavinginOperational3GCellularNetworks ACMMobicom2011LasVegas Nevada USA ChunyiPeng1 Suk BokLee1 SongwuLu1 HaiyunLuo HewuLi21UniversityofCalifornia LosAngeles2TsinghuaUniversity SurgingEnergyConsumptionin2G 3G 0 5 ofworld wideelectricitybycellularnetworksin2008 Fettweis 124Twhin2011 expected ABI CO2emission comparableto bycarsOperationcost e g 1 5BbyChinaMobilein2009Risingenergyconsumptionat16 20 yearMoore slaw 2xpowerevery4 5yearsby2030 Fettweis G FettweisandE Zimmermann ICTenergyconsumption trendsandchallenges WPMC 08 ABI ABIResearch Mobilenetworksgogreen minimizingpowerconsumptionandleveragingrenewableenergy 2008 EnergyConsumptioninCellularNetworks 0 1wX5B 0 5GW 1 3kwX4M 8GW 10kwX10K 0 1GW 90 99 CellularInfrastructure 10 1 MobileTerminals 80 byBSes Thekeytogreen3GisonBSnetwork Source NokiaSiemensNetworks NSN Outline OverviewProblemandrootcauseExistingsolutionsOursolutionCharacterizing3GdynamicsExploitingdynamicsindesignWorkingwith3GstandardsEvaluationSummaryandInsights CaseStudyinaRegional3GNetwork Non energy proportionality Non EP totrafficload RootCauseforEnergyInefficiency EachBSisnon EP Largeportionofconsumedenergyeven zerotrafficloadaslongastheBSison RootCauseforEnergyInefficiency TrafficishighlydynamicFluctuateovertimeBeunevenatBSes Largeenergyoverheadatlighttraffic non EP TurnoffBScompletelytosavemoreenergy Lowusageatnight GoalsandChallenges System wideenergyproportionality EP HowtodesignEPnetworkwithnon EPBScomponents NegligibleperformancedegradationHowtomeetlocation dep coverage capacityrequirements 3GstandardcomplianceHowtosupportenergyefficiencyw ochanging3Gstandard ExistingSolutions Optimization basedapproachPracticalissuesunaddressedTheoreticalanalysisonlyComponent basedapproache g oncooling poweramplifierNosystem widesolutionComplementourapproachCleanslatedesigne g C RANRe architectthe3GinfrastructureCommunicationandcomputationintensive subjecttoC1 C2 constraints OurSolutionRoadmap Atraffic drivenapproachthatexploitstrafficdynamicstoturnoffunder utilizedBSesforsystem wideenergyefficiencyCharacterizingmulti dimensionaldynamicsExploitingdynamicsindesignWorkingwith3GstandardsEvaluation TemporalDynamicsisPervasive LowaverageutilizationunderdynamicloadPeak to idletrafficis 5at40 80 BSes TemporalDynamicsisStable Temporalpatternisnear termstableTrafficateachBSisquitestableonadailybasisAutocorrelationwith24 hourlagis 0 92at70 BSesDay to dayvariation Curr Prev Prev is 0 2at70 BSes Autocorrelationwith24 hour lag SpatialDynamics DeploymentvariesatlocationsDenseinbigcities20 neighbor 1KM RichBSredundancyensurescoverage Spatial temporalDynamics TrafficisalsodiverseatvariouslocationsPeakhoursaredifferentMultiplexinggain 2atpeakhoursLowerboundfortheratioofcapacitytotraffic Multiplexinggain sum maxTraffic sum traffic Roadmap Characterizingmulti dimensionaldynamicsExploitingdynamicsindesignWorkingwith3GstandardsEvaluation IssueI HowtoSatisfyLocation dependentCoverage CapacityConstraints OnceaBSturnsoff clientsinitsoriginalcoverageshouldstillbecovered Evenifthetotalcapacityisenough itmayfailtoservemobileclientsduetocoverageissue providelocation dependentcapacity SolutionI BuildingVirtualGrids DivideintoBSvirtualgridsBSeswithinagridcovereachotherDecouplecoverageconstraintLocation dependentcapacitymeetslocation dep traffic VirtualBSGrids turnon offBSess t cap load j i ri d i j Rirj d i j Rj IssueII HowtoEstimateTrafficLoad Atwhattimescaleistrafficloadpredictable Exploitnearperiodicityoverconsecutivetime of the dayWhattoestimate Instantaneoustrafficloadvs trafficupper envelopeChoicesbetweenaccuracyandover estimateTradeoffbetweenenergyefficiencyandmiss rate SolutionII Profiling EstimatetrafficenvelopeviaprofilingLeveragenear termstabilityReduceruntimecomputation communicationReducemissrateviatrafficenvelopeestimation Stat Output IssueIII HowtoMinimizeOn OffSwitches Frequenton offswitchingisundesirableLargeramp uptimewhenonReducedlifetimeforcoolingandothersubsystemsHowoftentoswitchon off Over24 hourperiod consistentwithtrafficcharacteristics SolutionIII SmoothSwitches MonotonicallyincreasingONfromidle peakMonotonicOFFfrompeak idle AtmostONEon offswitchperBSper24hours Roadmap Characterizingmulti dimensionaldynamicsExploitingdynamicsindesignWorkingwith3GstandardEvaluation 2 Workingwith3GStandard Expand shrinkcoverageatONBsesCellbreathingtechniqueWhenneighborBSesturnOFF ON Triggernetwork controlledhandoffatOFFBSesLeveragehandoverproceduresBeforetheyturnoff 1 2 3 1 3 2OFF 1 2 3 CoordinateBSesatRNCviaIu binterfaceInformationcollectoranddistributor HowtoletONBSescoverthecomm areaofOFFBSes HowtomigrateclientsfromOFFBSestoONBSes Howtoshareinformationinavirtualgrid Roadmap Characterizingmulti dimensionaldynamicsExploitingdynamicsindesignWorkingwith3GstandardsEvaluation EnergySavinginFourRegions Usetwo monthrealtracesinfourregional3Gnetworks MoreonEvaluation OursolutionisrobusttovariousparametersettingsPowermodels capacity coverage profilingfactor Negativeimpactonclients MoreenergyforuplinkTxrange duetoON OFFschemeExampleinRegion1Negligibleatdaytime 1kmatnightCanbelessaggressive Summary ThecurrentcellularnetworkisnotenergyefficientItisfeasibletobuildapracticalsolutionto greencellularinfrastructure Especiallyint