Enabling网页计数器

NetworkGameResearch–IntroductionandCounter-strikeAnalysisMarkClaypoolOutlineOverviewofNetworkGamesNetGamesConferenceResearchIssuesAnalysisCounter-strikeOurResultsWhyStudyGamesRapidlyincreasinginpopularityForresterResearch:18millionon-linein2001Consoleson-linePlaystation2on-line(9/2002)XboxLive(12/2002)CellphonesDoomport(Nokia)DesignnetworkstobetteraccommodatetrafficGameTypesFirstPersonShootersDoom,Quake,Counter-strike,…MassiveMulti-PlayerOnlineRolePlayingEverquest,EarthandBeyond,…Real-TimeStrategyWarcraft,Starcraft…OtherMisc–notanygenreaboveExample:Diablo2,Racing,…Non-networkedExample:ThiefMultiplayer,butrelaxedrealtimeExample:Chess,BridgeNet-Games-WorkshoponNetworkandSystemSupportforGamesFirstwas2002inBraunschweig,Germanyhttp://wwwmc.
tm.
uni-karlsruhe.
de/netgames2002/SecondisinRedwoodCity,Californiahttp://confman.
eecs.
umich.
edu/netgames2003/Sponsoredby:ACMElectronicArtsMicrosoftNet-GamesProgramCommitteeSugihJamin,UniversityofMichigan(Chair)MostafaAmmar,GeorgiaInstituteofTechnologyGrenvilleArmitage,SwinburneUniversityofTechnologyJohnBuchanan,ElectronicArtsJonCrowcroft,UniversityofCambridgeChristopheDiot,SprintlabsWu-changFeng,OregonHealthandScienceUniversityCarstenGriwodz,UniversityofOsloJimKurose,UMassatAmherstJohnLaird,UniversityofMichiganBrianNeilLevine,UMassatAmherstMartinMauve,UniversityofMannheimHiroyukiMorikawa,UniversityofTokyoDanRubenstein,ColumbiaUniversitySrinivasanSeshan,CarnegieMellonUniversityWilsonYuen,CityUniversityofHongKongLarsWolf,TUBraunschweigNet-GamesTopicsMulti-playergamearchitecturesandplatformsPreventionanddetectionofcheatingGamesonmobileandresource-scarcedevicesAIandtechniquesforlatencyhidingModeling,usagestudies,andcharacterizationEnablingprotocolsfornetworkedgamesSystemssupportforauthenticationandaccountingPutresearchissueshereProvisioningOn-lineGames:ATrafficAnalysisofaBusyCounter-StrikeServerWu-changFeng,FrancisChang,Wu-chiFeng,JonathanWalpoleGoalUnderstandtheresourcerequirementsofapopularon-lineFPS(first-personshooter)gameWhyFPSWhilethereareothergametypes…Gamingtrafficdominatedbyfirst-personshootergenre(FPS)[McCreary00]NetworkedFPSlineageDoomDoomIIQuake+QuakeWorldvariants+TeamFortress+CapturetheFlagQuakeII+SoldierofFortune+HereticIIQuakeIIIArena+MedalofHonorAlliedAssault+ReturntoCastleWolfenstein+SoldierofFortune2+JediKnightIIDoomIIIHalf-Life+Counter-Strike+DayofDefeat+UrbanTerror+TeamFortressClassic+TeamFortress2UnrealUnrealTournamentUnrealTournament2003+America'sArmy:Operations8oftop10gamesderivedfromoneoftwolineagesAboutthegame.
.
.
Half-LifemodificationTwosquadsofplayerscompetinginroundslastingseveralminutesRoundsplayedonmapsthatarerotatedovertimeEachserversupportsupto32playersWhatisCounter-strike(CS)WhatisCounter-strike(CS)WhatisCounter-strike(CS)Aboutthegame.
.
.
CentralizedserverimplementationClientsupdateserverwithactionsfromplayersServermaintainsglobalinformationanddeterminesgamestateServerbroadcastsresultstoeachclientSourcesofnetworktrafficReal-timeactionandcoordinateinformationBroadcastin-gametextmessagingBroadcastin-gamevoicemessagingCustomizedsprayimagesfromplayersCustomizedsoundsandentiremapsfromserverWhyCSWhyCSTheTracecs.
mshmro.
com(129.
95.
50.
147)Dedicated1.
8GHzPentium4LinuxserverOC-370,000+uniqueplayers(WonIDs)overlast4monthsOneweekinduration4/11–4/18500millionpackets16,000+sessionsfrom5800+differentplayersAweekinthelife.
.
.
Variancetimeplot(Normalizedtobaseintervalof10ms)Lessvariance<50msRemainsto30minDecreasesabove30minDiggingdeeperIntervalsize=10msIntervalsize=50msPeriodicserverburstsevery50msGamemustsupporthighinteractivityGamelogicrequirespredictableupdatestoperformlagcompensationDiggingdeeperLowutilizationevery30minutesServerconfiguredtochangemapsevery30minutesTrafficevenlypeggedotherwise.
.
.
.
Intervalsize=1secIntervalsize=30minFindingthesourceofpredictabilityGamesmustbefairacrossallmediums(i.
e.
56kers)Aggregatepredictabilitydueto"saturationofthenarrowestlast-milelink"PacketsizesSupportingnarrowlast-milelinkswithahighdegreeofinteractivityrequiressmallpacketsClientssendsmallsingleupdatesServersaggregateandbroadcastlargerglobalupdatesImplicationsRouters,firewalls,etc.
mustbedesignedtohandlelargeburstsatmillisecondlevelsGamerequirementsdonotallowforlossordelay(lag)Shouldnotbeprovisionedassumingalargeaveragepacketsize[Partridge98]Iftherearebuffersanywhere,theymust.
.
.
UseECN(no,doesn'tuseTCP)Beshort(i.
e.
nothaveabandwidth-delayproductofbuffering)EmployanAQMthatworkswithshortqueuesRate-basednotQueue-basedImplicationsISPs,gameservicesMustexamine"lookup"utilizationinadditiontolinkutilizationConcentrateddeploymentsofgameserversmaybeproblematicLargeserverfarmsinasingleco-loAmerica'sArmy,UT2K3,XboxOn-goingworkOtherpiecesintheprovisioningpuzzleAggregateplayerpopulationsGeographicdistributionsofplayersovertime(IP2Geo)ImpactonrouteandpacketclassificationcachingOtherFPSgamesHL-based:DayofDefeatUT-based:UnrealTournament2003,America'sArmyQuake-based:MedalofHonor:AlliedAssaultResultsapplyacrossotherFPSgamesandcorroboratedbyotherstudiesFutureworkGamesaspassivemeasurementinfrastructureOnlywidespreadapplicationwithcontinuousin-bandpinginformationbeingdelivered(measurementforfree)"Pingtimes"ofallclientsbroadcasttoallotherclientsevery2-3seconds20,000+servers,millionsofclientsGamesasactivemeasurementinfrastructureThrivingFPSmodcommunityandtoolsServermodifications[Armitage01]OthergametypesSomeNet-GamesResearchatWPIMarkClaypool,DavidLaPoint,andJoshWinslow.
"NetworkAnalysisofCounter-strikeandStarcraft",InProceedingsofthe22ndIEEEInternationalPerformance,Computing,andCommunicationsConference(IPCCC),Phoenix,Arizona,USA,April2003.
Onlineat:http://www.
cs.
wpi.
edu/~claypool/papers/net-game/ComparedFPS(Counter-strike)withRTS(Starcraft)SmallernumberoftracesControlledsetofusersOurResultsConfirmspacketsizesforCounter-strikeStarcraftresults:PacketssmallerLinearincreasewithnumberofplayersStarcrafttrafficmuchsmoother(bandwidthusage)thanCounter-strikeCounter-strikeburstyduring'fire-fights'(PEDStalklater)NetworkGameResearch–IntroductionandCounter-strikeAnalaysisMarkClaypool