Size: px
Start display at page:

Download ""

Transcription

1 MeasuringWebPerformanceintheWideArea 111CummingtonSt,Boston,MA2215 PaulBarfordandMarkCrovella ComputerScienceDepartment BostonUniversity BU-CS-99-4 April26,1999 andcontinuedgrowthmakethisadicultquestiontoanswer.wedescribethewideareaweb whyusersoftenexperiencelongdelaysindocumentretrieval.theinternet'ssize,complexity, OneofthemostvexingquestionsfacingresearchersinterestedintheWorldWideWebis Abstract Measurementproject(WAWM)whichusesaninfrastructuredistributedacrosstheInternetto infrastructurearepresentedinthispaper. tracgeneratortocreaterepresentativeworkloadsonservers,andbothactiveandpassivetools studywebperformance.theinfrastructureenablessimultaneousmeasurementsofwebclient tomeasureperformancecharacteristics.initialresultsbasedonaprototypeinstallationofthe performance,networkperformanceandwebserverperformance.theinfrastructureusesaweb TheWorldWideWebisafascinatingexampleofadistributedsystemonanimmensescale.Its 1largescaleandincreasinglyimportantroleinsocietymakeitanimportantobjectofstudyand Introduction evaluation.atthesametime,sincethewebisnotcentrallyplannedorcongured,manybasic questionsaboutitsnaturearestillopen. preciselyaspossible,whilerecognizingthatitdoesnotadmittoasinglesimpleanswer. whataretherootcausesoflongresponsetimesintheweb?ourgoalistoanswerthisquestionas Thequestionthatmotivatesourworkisabasicone:whyistheWebsoslow?Moreprecisely: measurement.toanswerthisquestion,tworesearchneedsmustbeaddressed: 1.Theneedforintegratedserver/networkmeasurement.Mostrelatedresearchtodatehas AttemptingtoanswerthisquestionimmediatelyexposesgapsintheresearchtodateonInternet 2.Theneedtounderstandtherelationshipbetweenactiveandpassivemeasurementsinthe focusedonmeasuringserversandnetworksinisolation;theinteractionsbetweenthetwo, Internet.Activemeasurementsareeasilyunderstood,butdonotclearlypredictactualWeb especiallyinawide-areasetting,arenotwellunderstood. interpret. performance;passivewebmeasurementscanreectactualperformancebutcanbehardto 1

2 intendedtomeetthesetwoneeds.theprojectemploysbothnetworkandservermeasurement, andaswellusesbothpassiveandactivemeasurementapproaches.ourgoalinthispaperisto describetheproject,toexplainitsplaceinthecontextofinternetmeasurementresearch,andto Thispaperdescribesaprojectframework,calledWideAreaWebMeasurement(WAWM), demonstratetheutilityofourapproachusingsamplesofourearlymeasurements. taneously;ontakingconcurrentmeasurementsatboththeclientandtheserverendsofsample connections;andonvaryingserverloadsoastoallowcontrolledexplorationofawiderangeof server/networkinteractions.forexample,byvaryingserverloadandrunningexperimentsatdifferenttimesofday,wecanexplorethefourcasesinwhichtheserverandthewide-areanetwork Ourmeasurementarchitectureisbasedonmeasuringserverstateandnetworkstatesimul- issuesandactasaplatformfordevelopingtestinfrastructureandexpertise.itconsistsofaweb areindependentlyeitherheavilyorlightlyloaded(inqualitativeterms). theserver(evenuptoanoverloadedcondition).asingleo-siteclient,locatedacrosstheinternet serverrunningatoursite,alongwithanumberoflocalclientscapableofgeneratingvaryingloadon Thetestapparatusdecribedinthispaperisasmall-scaleprototypeintendedtoexposedesign (15hops)fromus,generatestestconnectionstoourserver.Passivepackettracemeasurementsare takenbothattheo-siteclientandattheserver.activemeasurementsoftheconnectionpathare alsocollectedduringeachtransferusingpoissonping(poip)[3]whichsendsudppacketsalong thesamepathastheconnectionandmeasuresthetime-averagednetworkstate(packetdelayand assessmentofwide-areawebperformance.nonethelesstheyyieldsomeimmediateinsightsabout deploy(asdescribedinsection3),itissucienttoyieldsurprisingandinformativeresults. lossrate).althoughthisapparatusisareducedversionofthefull-scalesystemwewilleventually andpassivenetworkmeasurement. thebenetsofintegratedserver/networkmeasurement,andabouttherelationshipbetweenactive TheresultswepresentinSection4areonlysamplesanddonotyetrepresentathorough serverloadistogenerateasignicantgapbetweenthetransmissionofconnectionsetuppacketsand therstdatapacketowingfromtheserver.whilethisgapcanbeunderstoodasaconsequence thenetwork.weshowthatforourexperimentalsetup,oneofthemostnoticeableeectsofhigh Forexample,wendthatserverloadhasdistinctiveeectsonthepatternofpacketowthrough ofsocketandleoperations,itisnotablethatfortypical(i.e.,short)transfers(2kb)thegap loaded(i.e.,packetlossratesarehigh),itisnotuncommonforaheavilyloadedservertoshow generallydominatestransfertime. bettermeanresponsetimethanalightlyloadedserver.ourmeasurementssuggestthatthismay bebecauseheavilyloadedserversoftenshowlowerpacketlossrates,andsincepacketlosseshave Inaddition,weshowanevenmoresurprisingeectofserverload.Whenthenetworkisheavily TCPconnections;thiseectmaybeduetothefactthatTCPpacketsowinaburstierpatternthan thatpoipmeasurementsofpacketdelayaregenerallylowerthanthoseexperiencedbypacketsin dramaticeectsontransferlatency,thiscanreducemeanresponsetime. dopoippackets.wealsoshowthatpoipmeasurementsofpacketlossarenotstronglypredictive Finally,wealsoexploretherelationshipbetweenactiveandpassivemeasurements.Weshow inthepresenceofloss. ofpacketlossexperiencedbytcpconnections,whichmaybeduetothefeedbacknatureoftcp themotivatingquestion:whyisthewebsoslow? thatalthoughtheanswersarenotyetinhand,ourapproachappearstoshowpromiseforaddressing Theseresultsdemonstratetheutilityofintegratednetwork/servermeasurement.Weconclude 2

3 2TheWAWMprojectisbasedonalargeanddiversebodyofpriorwork.Inparticular,werelyon resultsfromwebcharacterizationstudies,passiveandactivenetworkmeasurementstudies,time RelatedWork workwiththegoalsofthewawmproject. ofthestudiesfromeachoftheseareaswhichservedasthebasisforourworkandcontrastthat calibrationstudies,andwideareanetworkperformancestudies.inthissectionwepresentselections Alargebodyofworkhasdevelopedoverthepastveyearsthatpresentsmeasurementsand andperformancesuchasclientbehavior,serverbehavior,proxybehaviorornetworkbehavior. characterizationsoftheweb.thisworkhastypicallybeenfocusedonspecicaspectsofbehavior 2.1WebPerformanceCharacterization Clientstudiessuchas[2,24,25,33,39,72](aswellasclientproxystudies[2,16,26,31,42,45, 71,78,79])provideinformationonthebehaviorofusersacrossthemanyserversthattheymight accessduringbrowsingsessions.thisworkhasbeencriticalinbuildingthetoolsthatareusedto sametime.priorstudieshaveonlyhadaccesstoeithertheclientoraproxyortheserver. priorwebstudiesandthewawmprojectisthatwewillmonitorclientandserveractivityatthe simulateuseractivityinthewawmproject.themostsignicantdierencebetweenallofthe provideuswithmethodologiesformeasuringperformanceinboththeuserspaceandsystemspace. provideinsightintoserverbehaviorundervariousloadsandversionsofhttp.theyarefocusedon eithertheanalysisofhttplogsorondetailedmeasurementsofserverperformance.thesestudies Webserverbehaviorhasalsobeenstudiedextensively[5,6,11,14,15,32,47,55].Thesestudies aspossibleoratsomepredeterminedrate.workloadgeneratorssuchas[46,29,77]replayweb serverlogsorasummaryofserverlogs.weusethesurgeworkloadgenerator[1]togenerate generatorssuchas[1,9,23,22,59,75,76]areallbasedonmakingrepeatedrequestsasquickly ManytoolshavebeendevelopedforgeneratingworkloadsonWebservers.Syntheticworkload previousworkloadgeneratorsbecauseitincorporatesawiderangeorworkloadcharacteristicsthat representativelocalloadsonourserver.surgeisasyntheticworkloadgeneratorwhichisunlike areimportanttomanydierentaspectsofserverperformance. measurementsinoneplace(somewherenearanendpointofanhttptransaction)and,thus, tions.theshortfallofthesestudieswhenitcomestoanalyzinglatencyisthattheyonlytake showtheperformanceeectsofbothhttpprotocolinteractionandtcppacketlevelinterac- StudiesofthenetworkeectsofWebtracinclude[7,8,21,3,38,43,54,56].Thesestudies 2.2PassiveandActiveNetworkMeasurements cannotprovideinsightintothepacketdelaysinbothdirectionsofatransaction. Thetaskofassessinggeneralnetworkperformancehasbeenwellstudied.Agoodsourceofgeneral andthosewhichuseactivetechniques.passivetechniquestypicallyconsistofgatheringpacket informationoncurrentmeasurementtoolsandprojectscanbefoundat[73].generaltechniques leveldataatsometappointinthenetwork.examplesofpassivemonitorsincludecoral[58]and forstudyingnetworkperformancecanbesplitintothosewhichusepassivemonitoringtechniques in[17,18],andthecharacterizationoftheself-similarnatureoftracwasdevelopedfrompacket tracesin[41].anotherpassivemeasurementplatformhasbeenproposedin[44].thissystem, asinglepoint.forexample,earlypackettracestudiesofapplicationlevelinteractionsweredone tcpdump[74].agreatdealofinformationcanbeextractedfrompassivemonitoring,evenatonly calledwindmill,doesonlineanalysisofdatathathasbeencollectedpassively.oneofthebenets 3

4 passivemonitoringhaslimitationsintermsoftheamountofdatawhichisgatheredduringtests (typicallyverylarge)andthedicultyofrealtimeanalysisofthedata. ofpassivetechniquesisthattheirusedoesnotperturbthesystembeingmeasured.however, thenetworkandthenmeasuretheresponsetothatstimulus.activemeasurementtoolscanbe characteristicsbetweenhosts. usedtomeasurebulkthroughput,pathcharacteristics,packetdelay,packetloss,andbandwidth Activemeasurementtoolsinjectsomekindofstimulus(apacketorseriesofpackets)into thenetworkaccordingtoapoissonprocess.themotivationforthismethodisthatsamplingat intervalsdeterminedbyapoissonprocesswillresultinobservationsthatmatchthetime-averaged wecurrentlyusepoissonping(poip)[3]inthewawmproject.poipinjectsudppacketsinto Thereareanumberoftoolswhichcanbeusedtomeasureend-to-enddelayandpacketloss; stateofthesystem(assumingitisinastationarystate).thereareanumberofstudiesthathave usedsimilaractivemeasurementstostudypacketlossincluding[12,8]andtodrivemodelsoftcp approach. behaviorincluding[49,6].thesestudiesprovidebackgroundthatguidesouractivemeasurement \TracerouteReno"isatoolusedtomeasurebulkTCPthroughputcapacityoveranInternetpath. oute[36]isanactivemeasurementtoolthatgatherssimplepathcharacteristics.treno[48]or widthalongtheend-to-endpathincludingpathchar[37],bprobe[19],andnettimer[4].tracer- Anumberofactivemeasurementtoolshavebeendevelopedthatcanbeusedtomeasureband- projectwecompareactiveandpassivemeasurementsinordertounderstandhowactivemeasurementsofnetworkconditionscorrelatetoperformanceasseenattheclientandattheserver. datatheyinjectintothenetwork,theycanalsoperturbthenetwork'sperformance.inthewawm Theseactivemeasurementscangiveinsightintonetworkconditions,butdependingonhowmuch beendonebypaxsoninaseriesofstudiesreportedin[61,62,63,65].thesestudiesareastarting pointforthewawmprojectfrombothmeasurementandanalysisperspectives. Someofthemostextensiveworkusingbothpassiveandactivemeasurementtechniqueshas 2.3TimeCalibration Measurementsofend-to-enddelayinourarchitecturearebasedonhavingnearlysynchronized clocksontheclientandtheserver.thedicultiesofsynchronizingclocksinawidelydistributed Additionalstudiesofclocksynchronizationproblemsinclude[57,66].Paxsonpointsoutin[66] clientsandservers.ntpassuresthatclocksaresynchronizedontheorderoftensofmilliseconds. ofthenetworktimeprotocol(ntp)daemonwhichweuseinwawmtosynchronizetheclockson environmenthavebeenwellstudied,especiallybymills[52,53].thatworkledtothedevelopment thattheuseofntpprovidesclosesynchronizationofclocksbutthatchecksforskewshouldstillbe madewhenanalyzingdata.theparticularresultsreportedinthispaperarenotstronglysensitive 2.4WideAreaNetworkPerformanceMeasurement externaltimesource(gps). toclockskew,howeverinfutureworkweplantoeithercorrectforclockskeworuseasynchronized thepresenttime.noneproposetodothetypeofworkthatisbeingproposedforwawm;however Thereareanumberofwideareameasurementprojectswhichareeitherproposedorunderwayat therearesomesimilarities.theprojectswhicharethemostcloselyrelatedtowawmare: CooperativeAssociationforInternetDataAnalysis(CAIDA)[28].CAIDA'smissionisto \addressproblemsofinternettracmeasurementandperformanceandofinter-provider communicationandcooperationwithintheinternetserviceindustry." 4

5 TheNationalInternetMeasurementInfrastructure(NIMI)project[3].TheNIMIproject's IETF'sInternetProtocolPerformanceMetrics(IPPM)workinggroup[51,67].TheIPPM Internet." missionisto\createanarchitectureforfacilitatingameasurementinfrastructureforthe NLANR'sMeasurementandOperationsAnalysisTeam(MOATS)project[27].TheMOAT's workinggroup'smissionisto\developasetofstandardmetricsthatcanbeappliedtothe quality,performance,andreliabilityofinternetdatadeliveryservices." TheInternetPerformanceMeasurementandAnalysisproject(IPMA)[5].TheIPMA's missionisto\createanetworkanalysisinfrastructuretoderiveabetterunderstandingof missionisto\studytheperformanceofnetworksandnetworkingprotocolsinlocalandwideareanetworks." systemicservicemodelsandmetricsoftheinternet." KeynoteSystems,Inc.[34].KeynotePerspectiveisbilledasa\globalreal-timeservice" whichtheyuseto\continuallymeasuretheperformanceofpopularwebsitesandinternet TheSurveyorProject[69].EortsintheSurveyorprojectcenteronusingPoissonping-like backboneprovidersandregularlypublishtheresults(intheformofaperformanceindex)..." EachoftheseprojectsattemptstomeasureoranalyzesomeaspectofInternetperformance.The WAWMprojectissimilartotheaboveprojectsinthatitusesawideareameasurementinfrastructureandthattherearesignicantdatamanagementandanalysistasks.TheWAWMproject isdierentfromtheseforanumberofreasons.first,wawmprojectfocusesonauserlevelapplication theweb.mostoftheprojectslistedabovefocusongeneralnetworkmeasurement, tests,measurementsofuserlevelcharacteristicswillbemadeatthesametime.second,wawm proposestoinstallthesystemswhichwillactasserversintheinfrastructure.thisapproachfacili- monitoringandanalysis.althoughwawmproposestotakenetworklevelmeasurementsduring measurementsofone-waydelayandlossalongpathsbetweenanumberofremotesites. haveperformancemonitoringtoolswhichwillenabledetailedsystemmeasurementsduringtests. ofmanagement.third,thesystemswhichwillbeactingasserversinthewawmtestswillalso tateseaseandconsistencyofsoftwareinstallation,controlovertheenvironment,security,andease intofourcategories:measurementarchitectureandtools,managementtools,testingprotocol,and 3InthissectionwedescribethetoolsandmethodsusedintheWAWMproject.Webreakthemdown WAWMProjectOverview thespecicimplementationsusedintheresultsreportedhere. 3.1MeasurementArchitecture analysismethods.ineachcategorywerstdescribethelong-termprojectgoals,thenwedescribe 3.1.1GeneralArchitecture usedtoservedocumentstotheclients.serversarealsoconguredwithperformancemonitoring toolssothatdetailedmeasurementsofcpu,memoryandnetworkcharacteristicscanbemade ThegeneralarchitectureoftheWAWMhardwarecomponentsisshowninFigure1.Ontheleftare componentsthatarelocatedatoursite.therstsystemisconguredasthewebserverandisonly 5

6 Server Under Test LAN Load Generator(s) Figure1:GeneralArchitectureofWAWMHardwareComponents Active Passive Measurement Distributed Clients with thedurationofatest.thisisimportantsincewecannotexpectremotesystemstogenerate duringtests.thesecondandthirdsystemsareconguredaslocalloadgenerators.theyrunthe Surgeworkloadgeneratorwhichisusedtoadjusttheamountofloadontheserverthroughout Active Passive Measurement delays,packetlosses,pathroutes,andpathcharacteristicssuchasbottlenecklinkspeed.another systemisusedtotakebothpassivemeasurements(e.g.,packettraces)andactivemeasurements enoughloadtosaturatetheserverwhenwewanttoruntestsunderheavyserverload.thefourth system(notshown)isusedasthecontrolsystemwhichinitiatestestsandgathersdatafromthe infrastructure.allofthesystemsintheserverclusterareconnectedtothelocalinternetservice (e.g.,usingnetworkprobepackets).usingthesetoolsweexpecttomeasure(atminimum)packet istodistributethesesystemswidelyacrosstheinternetinanattempttoexploreawiderangeof pathtypes.theremotesystemsmakerequestsforlesfromtheserverduringtests.theyarealso provider(isp)suchthattheserverclusternetworktracisnotabottleneck. conguredwithbothpassiveandactivenetworkmeasurementtools.requestsaregeneratedby Ontherightsideofthegurearesystemsthatarelocatedremotelyfromoursite.Ourgoal clientsusingamodiedversionofsurge.thismodiedversioncanbeconguredtoruninits standardmodeortomakerepeatedrequestsforasingleleinmuchthesamewayaswebstone remotesiteswewouldliketohaveinthecompletewawm.thatstudyutilized37siteswhich [75]ẆeareinuencedbyPaxson'sNetworkProbeDaemonstudy[65]intermsofthenumberof enabledmeasurementofover1,distinctinternetpaths.selectionoftheadditionalsiteswillbe basedongeographiclocationandinternetconnectiontypewithheterogeneityasthegoal.atleast 3.1.2InitialPrototype oneadditionalserversitewillalsobeincludedtodiversifymeasurements. atbostonuniversity(bu)andconsistedofsixpcsconnectedviaswitched1mbpsethernet arealltakenfromthisprototypesystem.inourprototype,asingleserverclusterwaslocated issuesandactasaplatformfordevelopingtestinfrastructureexpertise.theresultsinthispaper Wehaveimplementedaprototypeofthisarchitectureinordertoexposemeasurementanddesign PentiumProCPUsand128MBofRAMeach.EachsystemranLinuxv2..3.Theserverwas conguredwithapache1.3.[68].traceroutev1.4a5andtcpdumpv3.4wereusedtomeasure routecharacteristicsandgatherpackettracesrespectively.xntpdv3-5.93wasusedtosynchronize whichisconnectedtothebuispviaa1mbpsbridge.thepcswereconguredwith2mhz timestampsbetweentheserverclusterandtheremotesystem.oneofthepcsintheservercluster 6

7 settoonesecondandwithapacketsizeof256bytes.thesystemsetupasthewebserverwas alsoconguredwithwebmonitor[4]tomeasuredetailedserverperformancecharacteristicsinthe wasconguredtorunbothtoandfromtheserverclusterwiththemeanmeasurementinterval wassetupasthetimeserverforalltests.thexntpdupdateintervalwassetto64seconds.poip CPU,memoryandnetworkcomponentsoftheserver. wererunundereitherloworhighlocalloadlevels.loadlevelsinsurgearedenedintermsof setuptomakerequestsforalesetwhichconsistedof2distinctles.thetotalsizeofthisdata setwasabout5mb;thustheentiredatasetcouldbecachedintheserver'sram.experiments ThelocalloadgeneratingsystemsintheserverclusterwereconguredwithSurgewhichwas arepresentedin[11].usinghttp/1.,weused4uesforlowloadand52uesforhighload. userequivalents(ues).characterizationsofloadunderarangeofuesforthesamepcsystems andranlinuxv2..35.itwasconnectedinthelocalareaviaswitched1mbpsethernetandthen ofdenver(du).thissystemwasconguredwitha333mhzpentiumprocpuand64mbofram totheduispviaa1mbpsbridge.thesystemwassetupwiththesamenetworkmeasurement TheprototypeinstallationusedonlyasingleremotesitewhichwaslocatedattheUniversity network:1kblesmeasureessentiallyonlytheconnectionsetuptime;2kblesaretypical-sized were1kb,2kband5kb.theselesizeswerechosentodemonstratearangeofbehaviorinthe andtimesynchronizationtoolsasthesystemsintheservercluster.themodiedversionofsurge Webtransfersandtypicallyinvolveabout6or7roundtrips,soaredominatedbyTCP'sslow-start wassetuptomakerepeatedrequestsforoneofthreelesontheserver.thethreelesizesselected 3.2ManagementArchitecture behavior;5kblesallowustoobservethemodeinwhichmostpacketsaresentintheweb, namelyintcp'scongestionavoidancemode. Inordertoruntests,downloaddataandmanagesystemswhicharepartoftheWAWMinfrastructure,asecuremanagementplatformisrequiredonallWAWMsystems.Weareevaluating third-partymanagementplatformsincludingthenimiplatformandinsoft'svitalagent[35].both providetheessentialmanagementfeaturesandexibilitynecessaryforwawmaswellasbuilt-in networkperformancemeasurementtools. designchoices.therefore,inordertogaininsightintotherelevantissues,wedevelopedourown managementtoolfortheprototypestudy.thetooliswritteninperlandisbasedonusingsecure Shell[7](SSHv1.2.26wasusedintheprototypetests).Thistoolprovedtobesucientforthe Themanagementplatformisanareaofthesysteminwhichpracticalmatterscandictate prototypetests. whichwawmtestsarerun.atthispointintimeasimplelestructureisadequatetostorethe data.eachtestruninourprototypestudywithonlyasingleclientresultedinabout9mbof byotherresearchers.thewawmdatarepositoryismaintainedatthesamecentralsitefrom Thedatacollectediscatalogedandmaintainedinsuchawayastofacilitateanalysisandreuse compressedresultsdata GeneralArchitecture 3.3TestingProtocol AWAWMtestisaperiodoftimeoverwhichaspecicsetofclientsystemsmakerequeststoa specicserver.specicationoftestsrequiresdeterminingthefollowing: 1.Theserverclusterwhichwillbeusedinthetest. 7

8 2.Thelocalloadwhichwillbeplacedontheserver. 3.Thesetofremoteclientswhichwillmakerequestsduringatest.Theinuenceofdistance 4.Thetimeofdaythetestwillberun.DuetothediurnalcycleofInternettracinNorth willallinuencetheoutcomeofmeasurements. fromtheserver,internetconnectiontypeandthenumberofclientsparticipatinginatest 5.Thedurationofthetest.Thetestmustberunlongenoughtocollectarepresentativesample America,testsrunduringthedaytypicallyexperiencemorepronouncedinuencefromcross ofdataforthegivennetworkandserverconditions. tracthanthoserunlateatnight. 6.Thescheduleforactivenetworkmeasurementsduringtests.Sinceactivemeasurementscould 3.3.2InitialPrototype perturbhttpperformance,thisschedulemustbeselectedsuchthattheresultingdatagives aclearpictureofthenetwork'sstatewhilenotinuencingthehttptests. Thetestingprotocolusedintheprototyperanonesetoftestsduringtheday(between11:amand 7:pm)andonesetoftestsatnight(between9:3pmand5:am)foreachoftendays.Initiating testsatthesetimesprovidedavarietyofnetworkconditionsforourdata.eachtestbeganby alwayssucienttotransferatleast4ofthe5kblesunderheavynetworkloadconditions routesduringthetests.eachtestwasrunfor15minutes;weobservedthatthisdurationwas runningtraceroutetoandfromtheserverwhichgaveusanindicationofthecurrentend-to-end and4696ofthe1kblesunderlightnetworkloadconditions.poipwasrunbothtoandfrom wererunduringeachtestset.theyconsistedofmakingrepeatedrequestsforasinglele(1kb, 2KBor5KB)underbothlowandhighserverloads. theserversimultaneouslywiththehttprequestsforthedurationofeachtest.sixseparatetests 3.4.1GeneralArchitecture 3.4AnalysisProtocol Oncetestdataiscollected,itisreducedandanalyzed.Automationoftheanalysisprocessis Ourmodelisbasedontcpanaly[64].Inadditiontostatisticalapproaches,weintendtodevelop necessaryduetothelargeamountofdatathatisgatheredduringeachtest.wehavebegunthe methodsfordetailedanalysisofindividualletransfers.detailedanalysismayrequireavisual developmentofatoolcalledtcpevalwhichwillperformthesetasksonthepackettracedata. \coordinationtool"muchlikethetooldevelopedin[13] InitialPrototype datagatheredthroughactiveandpassivemeasurements. Wealsousenatalie[3]toanalyzePoiptracesandintendtodevelopmethodstocorrelatethe Atthehighestlevelwecurrentlytabulatesummarystatisticsanddistributionalstatisticsofle ontimelinediagrams(e.g.,figures3,5,and7)becausetheycanrepresentthetimingofeventsat transferdelays,packetdelays,packetlossandserverperformance.tcpevalalsogeneratesanumber bothendsofaconnection. exchangesbetweenaclientandtheserverwithinasinglehttptransaction.inthispaperwefocus ofgraphicalaidstointerpretation,includingtimelinediagramswhichillustrateindividualpacket 8

9 Sizes FileServer LoadTrnsfdLatency Table1:Filetransferlatencystatisticsunderlightnetworkload Files MeanStd.Dev. Latency%pktlossmeandelaypktloss tcpdump tcpdumppoip%poipmean 5KB 2KB 1KB low delay KB 2KB 1KB high SampleResults.45 Inthissectionwereportoninitialresultsobtainedusingtheprototypeapparatusdescribedin measurementsshowedthattheroutetoandfromtheremotesitewasasymmetricbutfairlystable forallofthetests(13hopsfrombutoduand15hopsfromdutobu).whiletherewere Section3.Thetestingprotocolcalledfortraceroutetoberunbeforeandaftereachtest.These occasionalroutechangesbetweensetsoftest,wemeasuredonlyoneroutechangeduringoneof thetests. period,wepresentresultsofdatasetswhicharetypicalofallofthedatathatwascollected. networkloadandlesizesusedinthetests.whilemeasurementsweretakenoveratwoweek 4.1Measurementsofletransferlatency Firstwepresentperformancemeasurementsforletransfersovercombinationsofserverload, 4.1.1Performanceunderlightnetworkload Thedetailsoftheletransferlatencyunderlightnetworkloadforallthreelescanbeseenin Table1.Thetableshowsthatthebusyserveraddsapproximatelyonesecondtotheaveragele transferlatencyforthe1kband2kblesandover1.5secondstotheaverage5kbletransfer latency.meanpacketdelayandlosscharacteristicsunderlightnetworkloadasmeasuredbyboth PoipandtcpdumpcanbeseeninTable1.Delayandlossmeasurementsarefortheoutboundpath fromtheservertotheclient.thesenetworkmeasurementsshowthatthenetworkconditionswere fairlystablethroughoutthemeasurementperiod. relativelyconstantamountforeachlesize.theseeectscanbeseeinthecumulativedistribution forallthreelesizes.underheavyserverload,thevariabilityincreasesslightly.theprincipal eectofserverloadwhennetworkloadislowwastoincreasetheaverageletransfertimebya Underlightnetworkloadandlightserverload,theletransferlatencyhadverylowvariability percentageoftransfers,latencycanincreaseover6timeswhenserverloadishigh. function(cdf)diagramforthe2kbleinfigure2.thisdiagramalsoshowsthatforasmall networkmeasurementsisthroughtheuseoftcptimelineplots.thetimelinesfortypicaltransfers ofthe2kblecanbeseeninfigure3.intheseplots(aswellastherestofthetimelineplots inthissection),theclientisonthelefthandsideandtheserverisontheright.thetimeline Anotherwaytoanalyzeletransfersandtheeectsofvaryingserverloadforthelightlyloaded plotshighlightthedelayincurredforallletransferswhentheserverisbusy.afterthetcp 9

10 1.8.6 low server load high server load Figure2:CDFforletransferlatencyunderlightnetworkloadforthe2KBle lowserverload highserverload connectionsetupsequence,thereisroughlyaoneseconddelaybeforetherstdatapacketissent bytheserver.thisdelayisduetotheneedfortherequestfromtheclienttogetuptouserspace Figure3:Timelinediagramoftypical2KBletransfersunderlightnetworkload informationfromthecdfofletransferlatencyleadsustoconcludethatifthenetworkislightly loaded,themaximumadditionaldelaythatmosttransfersmightexperiencewhenaccessingour case,thisisnearlyinstantaneouswhentheserverislightlyloaded.thisdatacombinedwiththe andthenfortheresponsetogetbackdowntothenetwork.ascanbeseeninthelightlyloaded busyserverisaboutonesecond. Whenpacketdelaysandlossesincrease,transfercharacteristicschange.Thedetailsofthele transferlatencyunderheavynetworkloadforallthreelescanbeseenintable2.thenetwork 4.1.2Performanceunderheavynetworkload surementperiod.however,thesemeasurementsdonotcorrelatecloselywiththemeandelayand measurementsfrompoipindicatethatnetworkconditionswerefairlystablethroughoutthemea- networkloadtoheavynetworkload.theincreaseinpacketdelaysisanindicationofthelevelof lossvaluesextractedfromthetcpdumptraces.weexplorethisingreaterdetailinsection4.2. Packettransferdelaysextractedfromthetcpdumptracesincreasebetween13%and48%fromlight alsoshowsthatunderheavyserverloads,themeantransferlatencyincreasesinarangefrom1.3to increasesinarangefrom2.2to5timesthemeanlatencywhenthenetworkislightlyloaded.it congestioninthenetwork.thetableshowsthatunderlightserverloads,themeantransferlatency 1

11 Sizes FileServer LoadTrnsfdLatency Table2:Filetransferlatencystatisticsunderheavynetworkload Files MeanStd.Dev. Latency%pktlossmeandelaypktloss tcpdump tcpdumppoip%poipmean 5KB 2KB 1KB low delay.51 5KB 2KB 1KB high low server load.6 high server load.5 Figure4:CDFforletransferlatencyunderheavynetworkloadforthe2KBle whentheserverloadishighversuswhenitislow.thiseectisexploredinmoredetailinsection tonoteisthattransferlatencyofthe5kbleunderheavynetworkloadsactuallydecreases 2timesthemeanlatencywhenthenetworkislightlyloaded.Perhapsthemostinterestingstatistic takenduringthedaytimeperiodsforalldatasetsshowedhigherpacketlossanddelaycharacteristicsaswouldbeexpected.underheavynetworkload,letransferlatencyhadgenerallyhigher delayandhighervariabilityforallthreelesizesregardlessofserverload.theseeectscanbe Filetransferdelaysarestronglyinuencedbybothcongestionandpacketloss.Measurements infigure2,thekneeinthecdfoccuredaroundthe9thpercentile,infigure4,thekneeismuch seenbycomparingfigure2withthecorrespondinggureforheavynetworkload(figure4).while lowerindicatingthatamuchlargerfractionoftransfersexperiencedseveredelays. timesversusthelightlyloadednetworktimelines,whichisanindicationofthelevelofcongestion beseeninfigure5.asinthelightlyloadednetworkcase,thedelayinsendingtherstdata packetfromtheserverisevident.thesediagramsshowageneralelongationinpackettransfer Thetimelinesfortypicaltransfersofthe2KBlewhenthenetworkisheavilyloadedcan inthenetworkduringthemeasurement EectsofPacketLoss Measurementsmadeduringhighnetworkloadshowtheeectsofbothnetworkcongestionand packetloss.inthissectionwespecicallyexaminetheeectsofpacketlossonletransferlatency. 11

12 lowserverload highserverload Table3:Filetransferstatisticsunderheavynetworkloadforleswithnodroppedpacketsorat leastonedroppedpacketfor5kble Figure5:Timelinediagramoftypical2KBletransfersunderheavynetworkload Sizes FileServerFilesw/o Load pktlosslatency MeanStd.Dev.Fileswith Latency pktlosslatency MeanStd.Dev. 5KB 2KB 1KB low n/a n/a Latency KB 2KB 1KB high Thedetailsoftheeectsofpacketlossunderheavynetworkloadandhighserverloadcanbeseen isalsomuchlowerthanthatoflestransferredwithloss. transferredwithoutpacketloss.asmightbeexpected,variabilityforlestransferredwithoutloss Filetransfersforleswhichloseatleastonepacketarebetween1.3and7.8timesslowerthanles intable3.thistableshowsthedierencesbetweenlestransferredwithandwithoutpacketloss. 2KBle.Ascanbeseenfromthegure,packetlosscansignicantlyincreasebothmeanand variabilityofletransferdelay.whencomparingthisgurewithfigure4itcanbeseenthat almostalllongtransferdelaysareduetopacketdrops.onenotablefeatureofthegureisthe Figure6comparestheimpactofpacketlossonletransferlatencyforthe1KBleandthe secondtimeoutperiodforthere-sendingofalostsynpacketinthelinuxtcpimplementation. cuspataboutthe4%levelforthe1kble.thisisduetolostsynpackets.thereisathree torestartslowstartinordertotransfertheremainingpackets.theoveralleectofthislossisthe Figure7.Thisgureshowstheeectofasinglelostpacketduringthedatatransferphaseofthe HTTPtransactionunderlightserverloadandheavynetworkload.ThepacketlosswillcauseTCP Thetimelinesfortransfersofatypical2KBlewithandwithoutpacketlosscanbeseenin additionofabout.5secondstotheletransferlatency. 12

13 Figure6:CDFforletransferlatencyunderheavynetworkloadandheavyserverloadforles.4 withandwithoutpacketloss.3 1K without drops 2K without drops.2 1K with drops 2K with drops Figure7:Timelinediagramofa2KBletransfersunderheavynetworkloadandlowserverload withoutpacketloss withpacketloss

14 1.8 high server load low server load.6 Figure8:Filetransferlatencyforthe5KBletransfersunderheavynetworkload ever,closerexaminationofthedatafromalltendatasetsshowsaninterestingphenomenon.for Initialeectsofserverloadontransferlatencieshavebeenshownintheprecedingsections.How Eectsofserverload whenserverloadisincreasedduringthetests.furthermore,inthreeoutofthosesevencases,le transfersofthe5kble,insevenoutoftenofthedatasets,packetlossactuallygoesdown andthenetworkactivitycouldchangeenoughduringthistimetocauseperformancetoimprove. caseswhichisthesameasthedataintable2.thisresultcanpossiblybeattributedtooneoftwo things.first,theexperimentsforthe5kbletransferarerunapproximately45minutesapart latencydecreaseswhenserverloadisincreased.thiseectcanbeseeninfigure8foroneofthese otherpossibleexplanationisthattheheavyloadontheserverpreventsitfromproducingasmany However,activemeasurementsindicatethatthenetworkwasfairlyconsistentacrossalltests.An- burstsofpacketsasitmightnormallyifitwerelightlyloaded.weintendtofurtherinvestigate 4.2Activeversuspassivemeasurementsofnetworkconditions thiseectunderawiderrangeofconditionsandconclusivelydetermineitscause. thetcpstreamsthemselvesoftendisagree.thisdierenceisimportantbecauseitsuggeststhat Aswasseenintheprevioussection,valuesfordelayandlossfromPoipmeasurementsandfrom takenfromtcpstreams.infigure9weshowthecdfsofpacketdelayexperiencedbytwosetsof nections.inthissectionweexplorethesedierencesandtheirimplications. measurementsderivedfrompoipmaynotbeindicativeoftheexpectedperformanceoftcpcon- tolightlyloadednetworkconditions;theplotontherightcorrespondstoheavilyloadednetwork 5KBTCPowsandbyPoipduringthesametwotimeintervals.Theplotontheleftcorresponds OurrstobservationisthatPoip'smeasurementsofpacketdelayaretypicallylowerthanthose conditions.typicaldierencesbetweenthemeanvaluesreportedbypoipandthoseexperiencedby TCPpacketsareontheorderof1ms.NotethatthePoipmeasurementsweremadeconcurrently withthetcpconnections,sotheyaremeasuringthesamenetworkstate. thattcpoftentransmitspacketsinburstsontheorderofthecongestionwindowsize,thenwaits processisconsistentwiththenotionthattcp'spacketstreamisburstier(i.e.,interarrivalshave ahighercoecientofvariation)thanthatofapoissonprocess.thisseemsclearwhenconsidering ThehigherdelayexperiencedbyTCPpacketscomparedtopacketsarrivingviaaPoisson foracknowledgments.thestartofthenextburstisdeterminedbyroundtriptimeandgenerally occurssometimeafterthecurrentbursthasbeensent,leadingtoahighcoecientofvariationin 14

15 Poip high SL.5 Poip low SL Figure9:CDFofpacketdelayfromservertoclientduringtwo5KBletests(left:lightlyloaded.4 Pkts high SL.4 Pkts low SL network;right:heavilyloadednetwork;sl=serverload) Poip high SL Poip low SL Pkts high SL Pkts low SL 1 % Packet loss from Poip 8 6 Figure1:ScatterplotofpacketlossrateinTCPvs.Poip. 4 2 related.figure1showsascatterplotoflossrateexperiencedbytcp(onthexaxis)versusthat interpacketspacing. OursecondobservationisthatpacketlossasexperiencedbyTCPandbyPoiparenotstrongly % Packet from tcpdump experiencedbypoipduringthesametimeframe(ontheyaxis)forallofour5kbexperiments. Ifthetwomeasurementswerestronglyrelated,pointswouldtendtooccuraroundtheliney=x (whichisplottedforreference);theydonot.ingeneral,itseemsthatpoipseesahigherpacket lossestriggerfeedbackbehaviorintcp.uponencounteringpacketloss,tcpoftenentersastate lossratethantcp,althoughthisisnotuniversallytrue. inwhichnopacketsowforsometime,afterwhichitrecoversandpacketsbegintoowagainat areducedrate(forexample,seefigure7,rightside).thisbehaviormeansthattcp'sviewofthe Thereasonforthislackofcorrelationmaybethat unlikethecasewithpacketdelays packet fewerlosseventsthanareseenbyapoissonprocess(whichisindependentofpacketlossevents). networkisnotindependentofpacketlossevents;asaresult,itmaybethattcptendstoobserve ItseemsclearinanycasethatTCP'sfeedbackbehaviormeansthatitsviewofnetworklossesare conditionsasexperiencedbytcp.1asregardspacketdelays,theburstynatureoftcpows notstronglycorrelatedwiththatofatime-averageview. networkstate.itseemsthatpoiphasonlylimitedvalueinpredictingthenatureofnetwork 1Tobefair,itisimportanttonotethatPoipwasnotdesignedtoassessTCP'sviewofthenetwork,andthatit ThesetwoobservationshaveimplicationsfortheuseofPoipandsimilartoolstomeasure 15

16 meansthatpacketstypicallyseelargerqueuesthanthetime-averagedconditions.withrespectto packetloss,itseemsthatthefeedbacknatureoftcp'scongestionavoidancemakesitveryhard topredicttcp'sactualpacketlossesusinganopen-looptoollikepoip. 5InthispaperwehavedescribedtheWAWMprojectandarguedthatitrepresentsanapproach thatisbothnovelanduseful.itsnoveltyarisesfromitstreatmentoftheserverandnetworkas Conclusion behaviorandnetworkbehavior,andtoidentifycasesinwhichthetwodonotinteroperatewell. anintegratedsystem.itsutilitycomesfromtheabilitytoexploretheinteractionbetweenserver wehaveshownthatinmanyofourexperiments,serversunderhighloadsueredsignicantlyless canyieldinformativemeasurements.forexample,usingitwehaveshownthat(forourserver)the maineectofserverloadontypicaltransfersistodelaytherstdatapacketsent.inaddition Wehavedescribedasmall-scaleimplementationoftheprojectarchitectureandshownthatit packetlossthanthoseunderlowload.incomparingpacketdelayandlossesintcpconnections tomeasurementsobtainedwithpoipwendthatthereareconsiderable(andunderstandable) dierences.theseresultsareonlyinitiallooksatourdataandrequireconrmationinawider rangeofsettings;howevertheyaresuggestiveofinterestingeects. toexpandthemeasurementapparatusuntilitallowsustoassessrepresentativebehaviorofthe infrastructuretoaddressthisquestionwillrequireprogressalongtwodimensions.first,weneed Web.Thismeansthatweneedtoaddmoreremoteclients(frommorelocations),andconsidera Ourmotivatingquestioninthisprojectis:WhyistheWebsoslow?TousetheWAWM widerrangeofserverandplatforms(e.g.,iisrunningonwindowsnt). methodsthatoperateontracesofindividualtransfersandassesstherelativeimpactofserver hopetoobtainamorepreciseunderstandofthecausesoftransferlatencyintheweb. delay,networkdelay,packetloss,etc.ontransferlatency.usingthesemoredetailedanalyseswe Theseconddimensionisthatofanalyticmethods.Weintendtodevelopcharacterization studyaswellasprovidinginputontheprojectdesign.theauthorswouldalsoliketothanklars paper.theauthorswouldalsoliketothankvernpaxsonforprovidingpoissonpingforuseinthis Denverforaccesstothesystemwhichwasusedastheremoteclientforthetestspresentedinthis Acknowledgements.TheauthorswouldliketothankDavidMartinfromtheUniversityof Kellogg-Stedmanforhishelpinsettingupsystemswhichwereusedintheproject. References [1]WebBench2.. [2]G.Abdulla,E.Fox,andM.Abrams.Shareduserbehaviorontheworldwideweb.InWebNet97,Toronto, [4]J.Almeida,V.Almeida,andD.Yates.Measuringthebehaviorofaworldwidewebserver.InProceedingsof [3]A.Adams,J.Mahdavi,M.Mathis,andV.Paxson.Creatingascalablearchitectureforinternetmeasurement. Canada,October1997. [5]V.Almeida,A.Bestavros,M.Crovella,andA.deOliveira.CharacterizingreferencelocalityintheWWW. theseventhifipconferenceonhighperformancenetworking(hpn),whiteplains,ny,april1997. hasmanyusesotherthantheonethatweareconsidering. pages92{13,december1996. InProceedingsof1996InternationalConferenceonParallelandDistributedInformationSystems(PDIS'96), 16

17 [6]M.ArlittandC.Williamson.Internetwebservers:workloadcharacterizationandperformanceimplications. IEEE/ACMTransactionsonNetworking,5(5):631{645,October1997. [7]H.Balakrishnan,V.Padmanabhan,S.Seshan,M.Stemm,andR.Katz.Tcpbehaviorofabusyinternetserver: Analysisandimprovements.InProceedingsofIEEEINFOCOM'98,SanFrancisco,CA,March1998. [8]H.Balakrishnan,S.Seshan,M.Stemm,andR.Katz.Analyzingstabilityinwide-areanetworkperformance.In ProceedingsofACMSIGMETRICS'97,Seattle,WA,June1997. [9]G.BangaandP.Druschel.Measuringthecapacityofawebserver.InProceedingsoftheUSENIXAnnual TechnicalConference,Monterey,CA,December1997. [1]P.BarfordandM.Crovella.Generatingrepresentativeworkloadsfornetworkandserverperformanceevaluation. InProceedingsofACMSIGMETRICS'98,pages151{16,Madison,WI,June1998. [11]P.BarfordandM.Crovella.Aperformanceevaluationofhypertexttransferprotocols.InToappearin ProceedingsofACMSIGMETRICS'99,Atlanta,GA,May1999. [12]J.Bolot.End-to-endpacketdelayandlossbehaviorintheinternet.InProceedingsofACMSIGCOMM'93, SanFrancisco,Setpember1993. [13]L.Brakmo,S.O'Malley,andL.Peterson.Tcpvegas:Newtechniquesforcongestiondetectionandavoidance. InProceedingsofACMSIGMETRICS'96,Philadelphia,PA,May1996. [14]H.BraunandK.Clay.Webtraccharacterization:Anassessmentoftheimpactofcachingdocumentsfrom ncsa'swebserver.inproceedingsofthesecondinternationalwwwconference,chicago,il,october1994. [15]T.Bray.Measuringtheweb.InFifthInternationalWorldWideWebConference,Paris,France,May1996. [16]L.Breslau,P.Cao,L.Fan,G.Phillips,andS.Shenker.Webcachingandzipf-likedistributions:Evidenceand implications.inproceedingsofieeeinfocom'99,newyork,ny,march1999. [17]R.Caceres.Measurementsofwideareainternettrac.TechnicalReportUCB/CSD89/55,ComputerScience Department,UniversityofCalifornia,Berkeley,1989. [18]R.Caceres,P.Danzig,S.Jamin,andD.Mitzel.Characteristicsofwide-areatcp/ipconversations.InProceedings ofacmsigcomm'91,september1991. [19]R.CarterandM.Crovella.Measuringbottlenecklinkspeedinpacket-switchednetworks.InProceedingsof Performance'96,Lausanne,Switzerland,October1996. [2]L.CatledgeandJ.Pitkow.Characterizingbrowsingstrategiesintheworldwideweb.ComputerNetworksand ISDNSystems,26(6):165{173,1995. [21]S.Cheng,K.Lai,andM.Baker.AnalysisofHTTP/1.1onaWirelessNetwork.Technicalreport,Stanford University,1998. [22]HTTPClient. [23]TheStandardPerformanceEvaluationCorporation.Specweb96. [24]M.CrovellaandA.Bestavros.Self-similarityinworldwidewebtrac:Evidenceandpossiblecauses.InProceedingsofthe1996ACMSIGMETRICSInternationalConferenceonMeasurementandModelingofComputer Systems,Philadelphia,PA,May1996. [25]C.Cunha,A.Bestavros,andM.Crovella.Characteristicsofwwwclient-basedtraces.TechnicalReportTR-95-1,BostonUniversityDepartmentofComputerScience,April1995. [26]A.Feldmann,R.Caceres,F.Douglas,andM.Rabinovich.Performanceofwebproxycachinginhetrogeneous bandwidthenvironments.inproceedingsofieeeinfocom'99,newyork,ny,march1999. [27]NationalLaboratoryforAppliedNetworkResearch. [28]CooperativeAssociationforInternetDataAnalysis. [29]H.Frystyk-Nielsen.Libwww. [3]H.Frystyk-Nielsen,J.Gettys,A.Baird-Smith,E.Prud'hommeaux,H.Wium-Lie,andC.Lilley.Network performanceeectsofhttp/1.1,css1andpng.inproceedingsofacmsigcomm'97,cannes,france, Setpember1997. [31]S.Glassman.Acachingrelayfortheworldwideweb.ComputerNetworksandISDNSystems,27(2),1994. [32]S.GribbleandE.Brewer.Systemdesignissuesforinternetmiddlewareservices:Deductionsfromalargeclient trace.inproceedingsofthe1997usenixsymposiumoninternettechnologiesandsystems(usits'97), Monterey,CA,December

18 [33]B.Huberman,P.Pirolli,J.Pitkow,andR.Lukose.Strongregularitiesinworldwidewebsurng.Science, 28:95{97,1998. [34]KeynoteSystemsInc. [35]INSoft. [36]V.Jacobson.traceroute.ftp://ftp.ee.lbl.gov/traceroute.tar.Z,1989. [37]V.Jacobson.pathchar.ftp://ftp.ee.lbl.gov/pathchar/msri-talk.ps.gz,1997. [38]S.KhaunteandJ.Limb.Statisticalcharacterizationofaworldwidewebbrowsingsession.TechnicalReport GIT-CC-97-17,CollegeofComputing,GeorgiaInstituteofTechnology,1997. [39]T.Kwan,R.McGrath,andD.Reed.UseraccesspatternstoNCSA'sWWWserver.TechnicalReportUIUCDCS- R ,UniversityofIllinois,DepartmentofComputerScience,February1995. [4]K.LaiandM.Baker.Measuringbandwidth.InProceedingsofIEEEINFOCOM'99,NewYork,NY,March [41]W.Leland,M.Taqqu,W.Willinger,andD.Wilson.Ontheself-similarnatureofethernettrac(extended version).ieee/acmtransactionsonnetworking,pages2:1{15,1994. [42]B.LiuandE.Fox.Webtraclatency:Characteristicsandimplications.InWebNet98,Orlando,FL,November [43]B.Mah.AnempiricalmodelofHTTPnetworktrac.InProceedingsofINFOCOM'97,Kobe,Japan,April [44]R.MalanandF.Jahanian.Anextensibleprobearchitecturefornetworkprotocolperformancemeasurement. InProceedingsofACMSIGCOMM'98,Vancouver,Canada,September1998. [45]C.Maltzahn,K.Richardson,andD.Grunwald.Performanceissuesofenterpriselevelwebproxies.InProceedings ofacmsigmetrics'97,seattle,wa,june1997. [46]S.Manley,M.Courage,andM.Seltzer.Aself-scalingandself-conguringbenchmarkforwebservers.Harvard University,1997. [47]S.ManleyandM.Seltzer.Webfactsandfantasy.InProceedingsofthe1997USENIXSymposiumonInternet TechnologiesandSystems,Monterey,CA,December1997. [48]M.Mathis.Empiricalbulktransfercapacity.ftp://ftp.advanced.org/pub/IPPM/treno.txt,July1997. [49]M.Mathis,J.Semke,J.Mahdavi,andT.Ott.Themacroscopicbehaviorofthetcpcongestionavoidance algorithm.computercommunicationsreview,27(3),july1997. [5]InternetPerformanceMeasurementandAnalysisProjecct. [51]InternetProtocolPerformanceMetrics. [52]D.Mills.Networktimeprotocol(version3):Specication,implementationandanalysis.TechnicalReportRFC 135,NetworkInformationCenter,SRIInternational,MenloPark,CA,1992. [53]D.Mills.Improvedalgorithmsforsynchronizingcomputernetworkclocks.IEEE/ACMTransactionsonNetworking,3(3):245{254,June1998. [54]J.Mogul.Thecaseforpersistent-connectionHTTP.TechnicalReportWRL95/4,DECWesternResearch Laboratory,PaloAlto,CA,1995. [55]J.Mogul.Networkbehaviorofabusywebserveranditsclients.TechnicalReportWRL95/5,DECWestern ResearchLaboratory,PaloAlto,CA,1995. [56]J.Mogul,F.Douglis,A.Feldmann,andB.Krishnamurthy.Potentialbenetsofdeltaencodinganddata compressionforhttp.inproceedingsofacmsigcomm'97,cannes,france,setpember1997. [57]S.Moon,P.Skelley,andD.Towsley.Estimationandremovalofclockskewfromnetworkdelaymeasurements. InProceedingsofIEEEINFOCOM'99,NewYork,NY,March1999. [58]Coral:Passivenetworktracmonitoringandstatisticscollection. [59]UniversityofMinnesota.Gstoneversion1. [6]J.Padhye,V.Firoiu,D.Towsley,andJ.Kurose.Modelingtcpthroughput:Asimplemodelanditsempirical validation.inproceedingsofacmsigcomm'98,vancouver,canada,setpember1998. [61]V.Paxon.Fast,approximatesynthesisoffractionalgaussiannoiseforgeneratingself-similarnetworktrac. ComputerCommunicationsReview,27(5):5{18,October

19 [62]V.Paxson.Empirically-derivedanalyticmodelsofwide-areatcpconnections.IEEE/ACMTransactionson [64]V.Paxson.Automatedpackettraceanalysisoftcpimplementations.InProceedingsofACMSIGCOMM'97, [63]V.Paxson.End-to-endroutingbehaviorintheinternet.InProceedingsofACMSIGCOMM'96,PaloAlto,CA, August1996. Networking,pages316{336,August1994. [65]V.Paxson.End-to-endinternetpacketdynamics.InProceedingsofACMSIGCOMM'97,Cannes,France, [66]V.Paxson.Oncalibratingmeasurementsofpackettransittimes.InProceedingsofACMSIGMETRICS'98, Cannes,France,September1997. [67]V.Paxson,G.Almes,J.Mahdavi,andM.Mathis.FrameworkforIPperformancemetrics,RFC233. pages11{21,madison,wi,june1998. [7]SSHCommunicationsSecurity. [69]TheSurveyorProject. ippm=,1998. [68]ApacheHTTPServerProject. ftp://ftp.isi.edu/in-notes/rfc233.txt,1998. [72]L.Tauscher.Evaluatinghistorymechanisms:anempiricalstudyofreusepatternsinworldwidewebnavigation. [71]N.Smith.Whatcanarchivesoertheworldwideweb.InTheFirstInternationalWorldWideWebConference, Geneva,Switzerland,May1994. [75]G.TrentandM.Sake.Webstone:Therstgenerationinhttpserverbenchmarking,February1995.Silicon [73]CAIDAMeasurementToolTaxonomy. [74]tcpdump. Master'sthesis,DepartmentofComputerScience,UniversityofCalgary,Alberta,Canada,1996. [76]WebCompare. [78]R.Wooster.Optimizingresponsetime,ratherthanhitrates,ofwwwproxycaches.Master'sthesis,Virginia [77]Webjamma. GraphicsWhitePaper. [79]R.WoosterandM.Abrams.Proxycachingthatestimatespageloaddelays.InSixthFirstInternationalWorld [8]M.Yajnik,S.Moon,J.Kurose,andD.Towsley.Measurementandmodelingoftemporaldependenceinpacket WideWebConference,SantaClara,California,1997. PolytechnicInstituteandStateUniversity,Blacksbury,Virginia,1996. loss.inproceedingsofieeeinfocom'99,newyork,ny,march

Measuring Web Performance in the Wide Area Paul Barford and Mark Crovella Computer Science Department Boston University 111 Cummington St, Boston, MA 2215 fbarford,[email protected] BU-CS-99-4 April

More information

Application Level Congestion Control Enhancements in High BDP Networks. Anupama Sundaresan

Application Level Congestion Control Enhancements in High BDP Networks. Anupama Sundaresan Application Level Congestion Control Enhancements in High BDP Networks Anupama Sundaresan Organization Introduction Motivation Implementation Experiments and Results Conclusions 2 Developing a Grid service

More information

How To Monitor And Test An Ethernet Network On A Computer Or Network Card

How To Monitor And Test An Ethernet Network On A Computer Or Network Card 3. MONITORING AND TESTING THE ETHERNET NETWORK 3.1 Introduction The following parameters are covered by the Ethernet performance metrics: Latency (delay) the amount of time required for a frame to travel

More information

Testing the Cloud. By Dan Joe Barry

Testing the Cloud. By Dan Joe Barry Testing the Cloud By Dan Joe Barry Cloud computing has now passed from hype to reality. ore and more enterprises are realizing the benefits of remote hosting of IT services rather than local IT management,

More information

8x8 Network Monitoring Tool

8x8 Network Monitoring Tool 8x8 Version 1.0, May 2011 The Champion For Business Communications Table of Contents Introduction...3 Overview of 8x8 VoIP Readiness Tools...3 Test #1: VoIP Quality and Connectivity speed...3 Test #2:

More information

Packet Matching. Paul Offord, Advance7

Packet Matching. Paul Offord, Advance7 Packet Matching Paul Offord, Advance7 Relax! Model network Server Farm Client Router / Firewall Firewall Load Balancer LAN 1 Internet 0 2 3 4 5 The challenge Matching packets from PC to 1 st server tier

More information

Overview of Network Measurement Tools

Overview of Network Measurement Tools Overview of Network Measurement Tools Jon M. Dugan Energy Sciences Network Lawrence Berkeley National Laboratory NANOG 43, Brooklyn, NY June 1, 2008 Networking for the Future of Science

More information

VegaStream Information Note QOS Statistics

VegaStream Information Note QOS Statistics VegaStream Information Note QOS Statistics Why does the voice quality of this call, that sounded great yesterday, not sound so good today? or why do calls to this destination sound choppy? It is likely

More information

Transparent Optimization of Grid Server Selection with Real-Time Passive Network Measurements. Marcia Zangrilli and Bruce Lowekamp

Transparent Optimization of Grid Server Selection with Real-Time Passive Network Measurements. Marcia Zangrilli and Bruce Lowekamp Transparent Optimization of Grid Server Selection with Real-Time Passive Network Measurements Marcia Zangrilli and Bruce Lowekamp Overview Grid Services Grid resources modeled as services Define interface

More information

PES INSTITUTE OF TECHNOLOGY B.E 5TH SEMESTER (AUTONOMOUS) - PROVISIONAL RESULTS JANUARY 2015 COMPUTER SCIENCE AND ENGINEERING BRANCH

PES INSTITUTE OF TECHNOLOGY B.E 5TH SEMESTER (AUTONOMOUS) - PROVISIONAL RESULTS JANUARY 2015 COMPUTER SCIENCE AND ENGINEERING BRANCH 1 1PI12CS002 A A A A A B A A NA S NA NA NA NA NA NA NA 25.0 25.0 9.04 2 1PI12CS004 B I B C B A A A NA NA NA S NA NA NA NA NA 21.0 21.0 8.14 3 1PI12CS005 B B C B B B A B A NA NA NA NA NA NA NA NA 25.0 25.0

More information

Experimentation driven traffic monitoring and engineering research

Experimentation driven traffic monitoring and engineering research Experimentation driven traffic monitoring and engineering research Amir KRIFA ([email protected]) 11/20/09 ECODE FP7 Project 1 Outline i. Future directions of Internet traffic monitoring and engineering

More information

Introduction Page 2. Understanding Bandwidth Units Page 3. Internet Bandwidth V/s Download Speed Page 4. Optimum Utilization of Bandwidth Page 8

Introduction Page 2. Understanding Bandwidth Units Page 3. Internet Bandwidth V/s Download Speed Page 4. Optimum Utilization of Bandwidth Page 8 INDEX Introduction Page 2 Understanding Bandwidth Units Page 3 Internet Bandwidth V/s Download Speed Page 4 Factors Affecting Download Speed Page 5-7 Optimum Utilization of Bandwidth Page 8 Conclusion

More information

!NAVSEC':!A!Recommender!System!for!3D! Network!Security!Visualiza<ons!

!NAVSEC':!A!Recommender!System!for!3D! Network!Security!Visualiza<ons! !:!A!Recommender!System!for!3D! Network!Security!Visualiza

More information

Testing Packet Switched Network Performance of Mobile Wireless Networks IxChariot

Testing Packet Switched Network Performance of Mobile Wireless Networks IxChariot TEST PLAN Testing Packet Switched Network Performance of Mobile Wireless Networks IxChariot www.ixiacom.com 915-6649-01, 2006 Contents Testing Packet Switched Network Performance of Mobile Wireless Networks...3

More information

Intrusion Detection System

Intrusion Detection System Intrusion Detection System Time Machine Dynamic Application Detection 1 NIDS: two generic problems Attack identified But what happened in the past??? Application identification Only by port number! Yet

More information

CS551 End-to-End Internet Packet Dynamics [Paxson99b]

CS551 End-to-End Internet Packet Dynamics [Paxson99b] CS551 End-to-End Internet Packet Dynamics [Paxson99b] Bill Cheng http://merlot.usc.edu/cs551-f12 1 End-to-end Packet Dynamics How do you measure Internet performance? Why do people want to know? Are ISPs

More information

Technical Glossary from Frontier

Technical Glossary from Frontier Technical Glossary from Frontier A Analogue Lines: Single Analogue lines are generally usually used for faxes, single phone lines, modems, alarm lines or PDQ machines and are generally not connected to

More information

TCP tuning guide for distributed application on wide area networks 1.0 Introduction

TCP tuning guide for distributed application on wide area networks 1.0 Introduction TCP tuning guide for distributed application on wide area networks 1.0 Introduction Obtaining good TCP throughput across a wide area network usually requires some tuning. This is especially true in high-speed

More information

Cisco Performance Agent Data Source Configuration in the Branch-Office Router

Cisco Performance Agent Data Source Configuration in the Branch-Office Router Deployment Guide Cisco Performance Agent Figure 1. Application visibility in all network segments using Performance Agent in branch office Cisco Performance Agent is a licensed software feature of Cisco

More information

Monitoring WAAS Using Cisco Network Analysis Module. Information About NAM CHAPTER

Monitoring WAAS Using Cisco Network Analysis Module. Information About NAM CHAPTER CHAPTER 5 Monitoring WAAS Using Cisco Network Analysis Module This chapter describes Cisco Network Analysis Module (NAM), which you can use to monitor your WAAS devices. This chapter contains the following

More information

CIT 470: Advanced Network and System Administration. Topics. Performance Monitoring. Performance Monitoring

CIT 470: Advanced Network and System Administration. Topics. Performance Monitoring. Performance Monitoring CIT 470: Advanced Network and System Administration Performance Monitoring CIT 470: Advanced Network and System Administration Slide #1 Topics 1. Performance monitoring. 2. Performance tuning. 3. CPU 4.

More information

The ISP Column A monthly column on all things Internet

The ISP Column A monthly column on all things Internet The ISP Column A monthly column on all things Internet Just How Good are You? Measuring Network Performance February 2003 Geoff Huston If you are involved in the operation of an IP network, a question

More information

MEASURING WORKLOAD PERFORMANCE IS THE INFRASTRUCTURE A PROBLEM?

MEASURING WORKLOAD PERFORMANCE IS THE INFRASTRUCTURE A PROBLEM? MEASURING WORKLOAD PERFORMANCE IS THE INFRASTRUCTURE A PROBLEM? Ashutosh Shinde Performance Architect [email protected] Validating if the workload generated by the load generating tools is applied

More information

Measuring IP Performance. Geoff Huston Telstra

Measuring IP Performance. Geoff Huston Telstra Measuring IP Performance Geoff Huston Telstra What are you trying to measure? User experience Responsiveness Sustained Throughput Application performance quality Consistency Availability Network Behaviour

More information

Improving the Database Logging Performance of the Snort Network Intrusion Detection Sensor

Improving the Database Logging Performance of the Snort Network Intrusion Detection Sensor -0- Improving the Database Logging Performance of the Snort Network Intrusion Detection Sensor Lambert Schaelicke, Matthew R. Geiger, Curt J. Freeland Department of Computer Science and Engineering University

More information

Network Performance Measurement and Analysis

Network Performance Measurement and Analysis Network Performance Measurement and Analysis Outline Measurement Tools and Techniques Workload generation Analysis Basic statistics Queuing models Simulation CS 640 1 Measurement and Analysis Overview

More information

Lab VI Capturing and monitoring the network traffic

Lab VI Capturing and monitoring the network traffic Lab VI Capturing and monitoring the network traffic 1. Goals To gain general knowledge about the network analyzers and to understand their utility To learn how to use network traffic analyzer tools (Wireshark)

More information

An overview on Internet Measurement Methodologies, Techniques and Tools

An overview on Internet Measurement Methodologies, Techniques and Tools An overview on Internet Measurement Methodologies, Techniques and Tools AA 2012/2013 [email protected] (Agenda) Lezione 24/04/2013 Part 1 Intro basic concepts ISP Traffic exchange (peering)

More information

Comparison of Wireless Protocols. Paweł Ciepliński

Comparison of Wireless Protocols. Paweł Ciepliński Comparison of Wireless Protocols Comparison of Wireless Protocols Field test and comparing 82.11 protocol vs nstreme In Point To Multipoint scenarios IDEA? What force me to make such a comparison. Testing

More information

Date: 08/18/2015 Windows 2008R2 SP1 EndoWorks 7.4 Windows Updates Description Tested Pass/Fail Date

Date: 08/18/2015 Windows 2008R2 SP1 EndoWorks 7.4 Windows Updates Description Tested Pass/Fail Date Date: 08/18/2015 The following list of Microsoft Windows 2008R2 SP1 updates have been tested and approved for EndoWorks 7.4 compatibility. Prior to applying Windows Updates, make sure your system is current

More information

DATA$CENTER$FIREWALL$PRODUCT$ANALYSIS$$

DATA$CENTER$FIREWALL$PRODUCT$ANALYSIS$$ DATA$CENTER$FIREWALL$PRODUCT$ANALYSIS$$ $ $ Fortinet$FortiGate$1500D$v5.0,build0252 $ 2014$ $Ryan$Liles,$Chris$Thomas$ $ $ NSSLabs DataCenterFirewallProductAnalysis FortinetFortiGate1500D Overview NSSLabsperformedanindependenttestoftheFortinetFortiGate1500Dv5.0,build0252.Theproductwas

More information

GlobalSCAPE Wide Area File Services

GlobalSCAPE Wide Area File Services Wide Area File Services: Document Collaboration for the Distributed Business Environment The days of having everyone on a project together in the same office have long passed. To expand global reach and

More information

The next IP SLA generation Solution. Advisor SLA. Network Performance Monitoring Solution. www.hlog-qostelecom.com

The next IP SLA generation Solution. Advisor SLA. Network Performance Monitoring Solution. www.hlog-qostelecom.com The next IP SLA generation Solution Advisor SLA Network Performance Monitoring Solution 2 Advisor SLA Network Performance Monitoring Solution. Executive Summary The Advisor SLA Solution has been chosen

More information

EdgeMarc 4508T4/4508T4W Converged Networking Router

EdgeMarc 4508T4/4508T4W Converged Networking Router Introduction The EdgeMarc 4508T4W combines multiple voice and data features into a single, easy to use converged networking router. It includes models that have up to 4 T1 WAN interfaces or a single Ethernet

More information

Basic Multiplexing models. Computer Networks - Vassilis Tsaoussidis

Basic Multiplexing models. Computer Networks - Vassilis Tsaoussidis Basic Multiplexing models? Supermarket?? Computer Networks - Vassilis Tsaoussidis Schedule Where does statistical multiplexing differ from TDM and FDM Why are buffers necessary - what is their tradeoff,

More information

Super high-resolution video handling system and highly accurate video traffic monitoring technology. - Demonstrations at SC10 -

Super high-resolution video handling system and highly accurate video traffic monitoring technology. - Demonstrations at SC10 - Super high-resolution video handling system and highly accurate video traffic monitoring technology - Demonstrations at SC10-2010.11.16 NTT Laboratories. * This work is supported by the National Institute

More information

1.1. Abstract. 1.2. VPN Overview

1.1. Abstract. 1.2. VPN Overview 1.1. Abstract Traditionally organizations have designed their VPN networks using layer 2 WANs that provide emulated leased lines. In the last years a great variety of VPN technologies has appeared, making

More information

Part No. P0993139 05 March 24, 2004. Business Communications Manager. Call Detail Recording System Administration Guide

Part No. P0993139 05 March 24, 2004. Business Communications Manager. Call Detail Recording System Administration Guide Part No. P0993139 05 March 24, 2004 Business Communications Manager Call Detail Recording System Administration Guide 2 Copyright 2004 Nortel Networks All rights reserved. March 24, 2004. The information

More information

Network Architecture and Topology

Network Architecture and Topology 1. Introduction 2. Fundamentals and design principles 3. Network architecture and topology 4. Network control and signalling 5. Network components 5.1 links 5.2 switches and routers 6. End systems 7. End-to-end

More information

Electronic Health Records and Performance Metrics. Thomas E. Britten, Specialist Leader, Deloitte Consulting LLP

Electronic Health Records and Performance Metrics. Thomas E. Britten, Specialist Leader, Deloitte Consulting LLP Electronic Health Records and Performance Metrics. Thomas E. Britten, Specialist Leader, Deloitte Consulting LLP Electronic health records and performance metrics. Agenda Background Systems Engineering

More information

VoIP Application Note:

VoIP Application Note: VoIP Application Note: Configure NEC UX5000 w/ Nuvox SIP Trunking Service P/N 0913229 Date: 8/12/09 Table of Contents: GOAL... 3 PREREQUISITES... 3 SIP TRUNKING INFORMATION PROVIDED BY NUVOX... 3 NEC UX5000:...

More information

LOBSTER: Overview. LOBSTER: Large Scale Monitoring for Broadband Internet Infrastructure

LOBSTER: Overview. LOBSTER: Large Scale Monitoring for Broadband Internet Infrastructure LOBSTER: Overview LOBSTER: Large Scale Monitoring for Broadband Internet Infrastructure Herbert Bos* [email protected] http://www.cs.vu.nl/~herbertb Department of Computer Science Vrije Universiteit Amsterdam

More information

Cisco Application Networking for IBM WebSphere

Cisco Application Networking for IBM WebSphere Cisco Application Networking for IBM WebSphere Faster Downloads and Site Navigation, Less Bandwidth and Server Processing, and Greater Availability for Global Deployments What You Will Learn To address

More information

VoIP Application Note:

VoIP Application Note: VoIP Application Note: Configure NEC UX5000 w/ BroadVox SIP Trunking Service P/N 0913226 Date: 8/12/09 Table of Contents: GOAL... 3 PREREQUISITES... 3 SIP TRUNKING INFORMATION PROVIDED BY BROADVOX:...

More information

Broadband Quality of Service Experience (QoSE)

Broadband Quality of Service Experience (QoSE) Broadband Quality of Service Experience (QoSE) Indicators 1 Price is not the only dimension that is of interest to customers and regulators. Quality of Service Experience (QoSE) is integrally connected

More information

Packet Tracer - Subnetting Scenario 1 (Instructor Version)

Packet Tracer - Subnetting Scenario 1 (Instructor Version) (Instructor Version) Instructor Note: Red font color or gray highlights indicate text that appears in the instructor copy only. Optional activities are designed to enhance understanding and/or to provide

More information

Frequently Asked Questions

Frequently Asked Questions Frequently Asked Questions 1. Q: What is the Network Data Tunnel? A: Network Data Tunnel (NDT) is a software-based solution that accelerates data transfer in point-to-point or point-to-multipoint network

More information

The Ecosystem of Computer Networks. Ripe 46 Amsterdam, The Netherlands

The Ecosystem of Computer Networks. Ripe 46 Amsterdam, The Netherlands The Ecosystem of Computer Networks Ripe 46 Amsterdam, The Netherlands Silvia Veronese NetworkPhysics.com [email protected] September 2003 1 Agenda Today s IT challenges Introduction to Network

More information

How To Measure Quality Of Service On An Hpla (Html) Website

How To Measure Quality Of Service On An Hpla (Html) Website Relationship between Quality-of-Service and Quality-of-Experience for Public Internet Service Stas Khirman ([email protected]) and Peter Henriksen ([email protected]) Narus Inc. (http://www.narus.com) 395

More information

Ahsay Online Backup Suite v5.0. Whitepaper Backup speed

Ahsay Online Backup Suite v5.0. Whitepaper Backup speed Ahsay Online Backup Suite v5.0 Version 5.0.1.0 Jan 2006 Table of Content 1 Introduction...3 2 Testing Configuration and Setup...4 2.1 Hardware and Software Setup...4 2.2 Test Scenarios...4 3 Results...5

More information

Precision Time Protocol (PTP/IEEE-1588)

Precision Time Protocol (PTP/IEEE-1588) White Paper W H I T E P A P E R "Smarter Timing Solutions" Precision Time Protocol (PTP/IEEE-1588) The Precision Time Protocol, as defined in the IEEE-1588 standard, provides a method to precisely synchronize

More information

A Performance Analysis of Gateway-to-Gateway VPN on the Linux Platform

A Performance Analysis of Gateway-to-Gateway VPN on the Linux Platform A Performance Analysis of Gateway-to-Gateway VPN on the Linux Platform Peter Dulany, Chang Soo Kim, and James T. Yu [email protected], [email protected], [email protected] School of Computer Science,

More information

NETFORT LANGUARDIAN MONITORING WAN CONNECTIONS. How to monitor WAN connections with NetFort LANGuardian Aisling Brennan

NETFORT LANGUARDIAN MONITORING WAN CONNECTIONS. How to monitor WAN connections with NetFort LANGuardian Aisling Brennan NETFORT LANGUARDIAN MONITORING WAN CONNECTIONS How to monitor WAN connections with NetFort LANGuardian Aisling Brennan LANGuardian gives you the information you need to troubleshoot problems and monitor

More information

IP Networking Untethered

IP Networking Untethered IP Networking Untethered Alan O NeillO Flarion Technologies Reliable link Low delay Link Layer Wish List Far fewer end-to to-end retransmissions Better Efficient transport layer (TCP/IP) User Experience

More information

Thepurposeofahospitalinformationsystem(HIS)istomanagetheinformationthathealth

Thepurposeofahospitalinformationsystem(HIS)istomanagetheinformationthathealth FederatedDatabaseSystemsforReplicatingInformationin UniversityofDortmund,DepartmentofComputerScience,Informatik10 ExtendingtheSchemaArchitectureof E-mail:[email protected] HospitalInformationSystems

More information

Yealink VCS Network Deployment Solution

Yealink VCS Network Deployment Solution Yealink VCS Network Deployment Solution Feb. 2015 V10.15 Yealink Network Deployment Solution Table of Contents Table of Contents... iii Network Requirements Overview... 1 Bandwidth Requirements... 1 Bandwidth

More information

IEEE Congestion Management Presentation for IEEE Congestion Management Study Group

IEEE Congestion Management Presentation for IEEE Congestion Management Study Group IEEE Congestion Management Presentation for IEEE Congestion Management Study Group Contributors Jeff Lynch IBM Gopal Hegde -- Intel 2 Outline Problem Statement Types of Traffic & Typical Usage Models Traffic

More information

F5 Configuring BIG-IP Local Traffic Manager (LTM) - V11. Description

F5 Configuring BIG-IP Local Traffic Manager (LTM) - V11. Description F5 Configuring BIG-IP Local Traffic Manager (LTM) - V11 Description This four-day course gives networking professionals a functional understanding of the BIG-IP LTM v11 system as it is commonly used, as

More information

SIP Trunking Service Configuration Guide for Skype

SIP Trunking Service Configuration Guide for Skype Notice Note that when converting this document from its original format to a.pdf file, some minor font and format changes may occur. When viewing and printing this document, we cannot guarantee that your

More information

Lab 1: Evaluating Internet Connection Choices for a Small Home PC Network

Lab 1: Evaluating Internet Connection Choices for a Small Home PC Network Lab 1: Evaluating Internet Connection Choices for a Small Home PC Network Objective This lab teaches the basics of using OPNET IT Guru. We investigate application performance and capacity planning, by

More information

STI Hardware Specifications for PCs

STI Hardware Specifications for PCs Local School Fileserver with STIOffice & STIClassroom WIN 1GHz or Higher 512MB 1GB Free Space Windows 2000/2003 Server, Novell Netware v4.11 or greater. Local School Fileserver with STIOffice, STIClassroom

More information

1 Analysis of HTTPè1.1 Performance on a Wireless Network Stephen Cheng Kevin Lai Mary Baker fstephenc, laik, [email protected] http:èèmosquitonet.stanford.edu Techical Report: CSL-TR-99-778 February

More information

Network Layer IPv4. Dr. Sanjay P. Ahuja, Ph.D. Fidelity National Financial Distinguished Professor of CIS. School of Computing, UNF

Network Layer IPv4. Dr. Sanjay P. Ahuja, Ph.D. Fidelity National Financial Distinguished Professor of CIS. School of Computing, UNF Network Layer IPv4 Dr. Sanjay P. Ahuja, Ph.D. Fidelity National Financial Distinguished Professor of CIS School of Computing, UNF IPv4 Internet Protocol (IP) is the glue that holds the Internet together.

More information

Terms & Conditions Server Hosting

Terms & Conditions Server Hosting Terms & Conditions Server Hosting Contents 1. Charging and Invoices... 2 1.1 Set up charges... 2 1.2 Invoicing... 2 2. Netnorth Service Demarcation Diagram... 2 2.1 NIC... 2 2.2 NF... 3 2.3 Service Demarcation...

More information

Sangoma Hardware Transcoder Troubleshooting/Debugging

Sangoma Hardware Transcoder Troubleshooting/Debugging Sangoma Hardware Transcoder Troubleshooting/Debugging The following document will help you troubleshoot possible issues with your Transcoder To begin troubleshooting please follow these steps in order:

More information

Information Security Training. Assignment 1 Networking

Information Security Training. Assignment 1 Networking Information Security Training Assignment 1 Networking By Justin C. Klein Keane September 28, 2012 Assignment 1 For this assignment you will utilize several networking utilities

More information

WIRELESS BANDWIDTH MANAGEMENT AUTHENTICATION IMPROVING QUALITY OF SERVICE

WIRELESS BANDWIDTH MANAGEMENT AUTHENTICATION IMPROVING QUALITY OF SERVICE WIRELESS BANDWIDTH MANAGEMENT AUTHENTICATION IMPROVING QUALITY OF SERVICE Amanda PEART & Alice GOOD ABSTRACT: With the popularity of distributed applications such as BitTorrent and Peer 2 Peer (P2P) networks,

More information

Bandwidth Security and QoS Considerations

Bandwidth Security and QoS Considerations This chapter presents some design considerations for provisioning network bandwidth, providing security and access to corporate data stores, and ensuring Quality of Service (QoS) for Unified CCX applications.

More information

Introduction to Performance Measurements and Monitoring. Vinayak Hegde [email protected] @vinayakh

Introduction to Performance Measurements and Monitoring. Vinayak Hegde vinayakh@gmail.com @vinayakh Introduction to Performance Measurements and Monitoring Vinayak Hegde [email protected] @vinayakh 1 Structure Part 1 Basics of Measurement (40 min) Part 2 Basic Statistics primer (40 min) Part 3 Measurement

More information

Optimum Business SIP Trunk Set-up Guide

Optimum Business SIP Trunk Set-up Guide Optimum Business SIP Trunk Set-up Guide For use with IP PBX only. SIPSetup 07.13 FOR USE WITH IP PBX ONLY Important: If your PBX is configured to use a PRI connection, do not use this guide. If you need

More information

IHM VoIP Products. Document history:

IHM VoIP Products. Document history: IHM P/S Vandtaarnsvej 87 DK-2860 Soeborg Denmark IHM VoIP Products Document history: Version: Description: 1.00 24-04-2015 EH First Version 1.01 29-04-2015 SBS Images included and minor text changes made.

More information

BT Internet Connect Global - Annex to the General Service Schedule

BT Internet Connect Global - Annex to the General Service Schedule 1. Definitions The following definitions apply, in addition to those in the General Terms and Conditions and the General Services Schedule. ARP means Address Resolution Protocol. Border Gateway Protocol

More information

Network monitoring in high-speed links

Network monitoring in high-speed links Network monitoring in high-speed links Algorithms and challenges Pere Barlet-Ros Advanced Broadband Communications Center (CCABA) Universitat Politècnica de Catalunya (UPC) http://monitoring.ccaba.upc.edu

More information

Scenario 1: One-pair VPN Trunk

Scenario 1: One-pair VPN Trunk VPN Trunk Load-Balance between Vigor3200 and Other Vigor Router This section will discuss how to build VPN Trunk with load-balance between Vigor3200 and other router (e.g., Vigor3300). Scenario 1: One-pair

More information

HTGR- Netflow. or, how to know what your network really did without going broke

HTGR- Netflow. or, how to know what your network really did without going broke HTGR- Netflow or, how to know what your network really did without going broke Michael W. Lucas [email protected] GKN Driveline North America, Inc. Copyright 2007 Michael W. Lucas slide 1 What

More information

CSE 473 Introduction to Computer Networks. Exam 2 Solutions. Your name: 10/31/2013

CSE 473 Introduction to Computer Networks. Exam 2 Solutions. Your name: 10/31/2013 CSE 473 Introduction to Computer Networks Jon Turner Exam Solutions Your name: 0/3/03. (0 points). Consider a circular DHT with 7 nodes numbered 0,,...,6, where the nodes cache key-values pairs for 60

More information

1000Mbps Ethernet Performance Test Report 2014.4

1000Mbps Ethernet Performance Test Report 2014.4 1000Mbps Ethernet Performance Test Report 2014.4 Test Setup: Test Equipment Used: Lenovo ThinkPad T420 Laptop Intel Core i5-2540m CPU - 2.60 GHz 4GB DDR3 Memory Intel 82579LM Gigabit Ethernet Adapter CentOS

More information

Hosted Voice. Best Practice Recommendations for VoIP Deployments

Hosted Voice. Best Practice Recommendations for VoIP Deployments Hosted Voice Best Practice Recommendations for VoIP Deployments Thank you for choosing EarthLink! EarthLinks best in class Hosted Voice phone service allows you to deploy phones anywhere with a Broadband

More information

Lab 4.1.2 Characterizing Network Applications

Lab 4.1.2 Characterizing Network Applications Lab 4.1.2 Characterizing Network Applications Objective Device Designation Device Name Address Subnet Mask Discovery Server Business Services 172.17.1.1 255.255.0.0 R1 FC-CPE-1 Fa0/1 172.17.0.1 Fa0/0 10.0.0.1

More information

International Telecommunication Union. Common VoIP Metrics. Alan Clark. CEO, Telchemy

International Telecommunication Union. Common VoIP Metrics. Alan Clark. CEO, Telchemy International Telecommunication Union Common VoIP Metrics Alan Clark CEO, Telchemy Workshop on End-to-End Quality of Service.What is it? How do we get it? Geneva, 1-3 October 2003 Summary ITU-T o Typical

More information

An Ethernet based alarm system

An Ethernet based alarm system http://tuxgraphics.org/electronics An Ethernet based alarm system Abstract: Many alarm system for buildings use dedicated wiring. They are often just alarming and monitoring one single building. Using

More information