中国移动GPRS无线优化专业培训资料

中国移动GPRS无线优化课题交流会SolutionCenterGTSSChinaMotorola1.GPRS业务模型分析MotorolaUSTMViewofUserServicesUSTMconsidersserviceactivityfromtheGPRS/UMTSuser’spointofview:(USTMstandsforUnifiedSimpleTrafficModels)UserApplicationTelecommunicationsNetworkUserApplicationApplicationLevelServiceActivityAtthisleveltheapplicationservicebehaviourisrelativelynetworkimplementationindependent.However,asthetrafficistransportedoverthenetwork,trafficbehaviourwithinthenetworkwilldependonthenetworkimplementationandconfigurationExample:Application<->ProtocolsInteractionAstheapplicationinteractswithdifferentprotocolsinthenetworkimplementation,applicationtrafficbecomespackets.ApplicationApplicationUMTSNetworkUTRANCoreNetworkInternetTCP/IPPacketActivityApplication-levelActivityPacketbehaviourmaybedifferentindifferentprotocollayers,andpartsofthenetwork,i.e:UTRANUMTSCoreNetworkExternalPacketDataNetworkTCP&IPpacketsRLCSDURadioframesLayeredTrafficModellingApproachThetrafficcharacteristicsofaservicecanbeverycomplex.Itisoftendescribedbyalayeredserviceworkloadmodel–dataapplicationmodelinUSTM.ThefollowingisaviewoftheFTPservice:SessionBurst(File)PackettttDefinitionofASessionAservicesessionisalogicalentity.Independentoftransmissiondirection.Itisbasicallydescribedastheperiodwhereauserisactivelyusingtheservice.Duringanactiveservicesession,thenetworkmustmaintainconnectivitybetweentheusersinthatsession.Service

DefinitionofSessionTelephony: TelephonecalldurationWWW: Webbrowsingsession(difficulttodeterminefinish)FTP: PerioddownloadingfromsingleserverEmail(download): FromconnectingtoemailservertodisconnectionSessionsaredescribedby:SessionDuration(D)SessionInter-arrivaltime(I)DItDefinitionofABurstService

ExampleofBurstTelephony: ActivetalkspurtWWW: HTTPobjectsinaWWWpagedownloadFTP: Transferofafilebetween

serverclientEmail(download): AfilecontainingsingleemailwithMIMEencodedattachmentsSMS: OnetextmessageBurstsdescribedby:BurstSize/Duration()Inter-bursttime(T)TtTheBurstrepresentsablockofrealtrafficthattheapplicationsends/receives.TheBurstistheactivitythatthebearerchannelsinthenetworksee.Formanyservices,aburstequatestothetransferofafileormessage.Theburstisuni-directional(uplinkanddownlinkburstsaredifferent).Note:Burstdurationmaybedrivenbydatasize/linkspeed/loadMoreExamplesofBurstTelephony:SpeechburstsUplinkDownlinkWWW:FileburstsUplinkDownlinkPageDownloadsSMSMO:MessageBurstUplinkSpeechsamplesencodedintospeechframes.Whilstoriginatoristalking,speechframeactivitygeneratesthespeechburst.ComplexWWWpagesmaycontainmultipleobjectslocatedondifferentservers.DownloadinganobjectisanHTTPrequest/responsetransaction.UnidirectionaltransmissionoftheSMSpacket.AcompleteView–Example:WWWTrafficModelSAWWWApplicationSummary:SessionObjectstDItBurststULDLBurstPacketTD - SessiondurationI - Sessioninter-arrivaltime - BurstdurationT - Inter-bursttimeS - PacketsizeA - Packetinter-arrivaltimePDPContextsAPDPcontextisalogicalpipethatprovidesthetransferofuserdataacrosstheUMTSnetwork.UMTSExternalNetworkApplicationPDPContextApplicationFromeitherend,thePDPcontextwilllooklikeanIPtermination(IPaddress/port)thatoffersagivenlevelofIPQoS(e.g.actsasaDiffServcodepoint).Note:ThePDPcontextdoesnotreallyseetheuserapplication,butonlyseestheIPQoSrequirements.UMTSUserLANPDPExt.IPTETETETEUMTSlookslikeanIProuterRealNetwork=MSModemGGSNUEUERemoteHostPDPContextsvs.SessionsinUSTMAservicesessionneedsaconnectionacrosstheIPnetworkstooperate.ThusaservicesessionneedsaPDPcontextoverUMTSforitsduration.However,aPDPcontextmayexistbeforeandafteraparticularservicesession.MorethanoneservicesessioncanuseasinglePDPcontext.DifferentservicessessionsmayshareasinglePDPcontext.PDPcontextsmaychangestates(active,standby).Thisistransparenttotheservice.ServiceAsessionsServiceBsessionsPDPContextstDataCollectionpointsRANCorePDNGbMDTTPCtcpdump/windumpTcptrace/EtherrealNetsenseGiGbConsultantRadComGbLinkAnalysisEnd-to-EndAnalysisPerformanceBenchmarkingtcpdump/windumpTcptrace/EtherrealNetsenseEnd-to-EndAnalysisEndUser(DataandRFStats)K1205GbSnoopGDSScriptsBOTGiLinkAnalysistcpdumpWindumpTcptraceEtherrealNetsenseTCPAnalyzerTEMS用户数据流与协议栈用户数据(IP包)从终端出发跨过GPRS承载到达因特网,途中进行了大量的协议封装和解封装,传输完成后,在GGSN还原成原来的用户数据(IP包),从GGSN经Gi接口进入因特网,我们选择Gi接口作为用户数据流的侦听和分析点.BSS/PCUSGSN因特网GGSNPPPIPTCP/UDPHTTPetc.L2L1IPTCP/UDPHTTPetc.L2L1Gi接口用户数据的封装WAP数据从GGSN出发封装到GRE隧道协议中去WAP网关,WAP网关作为代理访问服务站点WWW服务,电子邮件服务以及其它因特网服务经过TCP/UDP封装后,直接进入因特网去服务站点因特网GGSNWapGWGRETCP/UDPwwwserverOtherAppsMailserviceIPTCP/UDP/GREHTTP/WAP/POP3&SMTP/etc.L2L1NetworkTransportApplicationDataLinkPhysical协议栈与分析内容ApplicationPresentationSessionTransportNetworkDataLinkPhysicalApplicationPresentationSessionTransportNetworkDataLinkPhysical封装后的应用层报文和实际传输的数据,如FTPDataApplicationApplication应用层报文(控制消息等),如FTPControl在应用层只能看到“应用消息”,它被封装到传输层和网络层与数据一起被传输,在应用层的话务量分析没有意义.应用层做会话分析,传输层做连接分析和话务分析.传输协议分析应用协议分析应用会话分析传输连接分析传输流量分析网络流量分析分析内容分析内容CaseStudyin2003fromaEuropeanNetworkDetailedAnalysis–General

TotalFiguresTotalFigures AllSessions +10kByteSessionsTotalTraffic: 440MByte 330MByteDL/ULRatio: 4,75 32,55Sessions: 64778 6936TCPSessions: 80% 100%SessionDuration(Avg): <10s(85%) 1min50secTrafficperTCPSession(Avg): 7,1kB 47kBTrafficperUDPSession(Avg): 1,1kB N/AAverageSegmentSize(DL/UL): 320/N/ABytes 780/330Bytes+10kByteSessionsAllsessionswithmorethan10kBytetrafficindownlinkarecounted.Source:TCPDump/TCPTraceDetailedAnalysis–General

ApplicationUsage(intermofsessions)Source:TCPDump/TCPTraceAllSessions1%~650Sessions+10kBSessions(inDL)1%~70SessionsDetailedAnalysis–General

ApplicationUsage(intermofbytes)Source:TCPDump/TCPTraceAllSessions1%~650Sessions+10kBSessions(inDL)1%~70SessionsDetailedAnalysis–General

TransportProtocolUsageMajorTrafficTCPHTTPEmailSource:TCPDump/TCPTraceUDPWAPDNSBytesSessionsAllSessionsDetailedAnalysis–General

PacketSize(ULandDL)Source:TCPDump/NetSenseAllSessionsDetailedAnalysis–TCPParameterSegmentSizeRequestedAppliedWindowSizeInitial,Avg/StdDev/Min/MaxAppliedSelectiveACKUsageOutofOrderPacketsSource:TCPDump/TCPTraceDetailedAnalysis–TCPParameter

SelectiveAcknowledge/OutofOrder(onGi)Source:TCPDump/TCPTraceSelectiveAcknowledge AllSessions +10kByteSessions(DL)DL: 24% 0,04% UsedinX%oftheTCPSessionsUL: 24% 2% UsedinX%oftheTCPSessionsOutofOrderPacketsDL: 0,16% 2,93% TCPSessionswithOoOPacketsUL: 0,07% 0,22% TCPSessionswithOoOPacketsConclusion&RecommendationsSystemOptimisationHW/SWDimensioningFeaturePriorizationConclusion&Recommendations

SystemOptimisationShortSessions-GeneralHTTP1.1isusedCausesLowThroughputLowSACKusageforlargetransfers3%outoforderpacketforlargetransfersonGi!Recommendation:TCPStackTuningProxySelectiveACK,RetransmissionsTCPClientWindowSizeTuning(viaDriverInstallation?)ImportantforCS3-4FurtherAnalysison„Whysomanyshort(HTTP)sessions?“Conclusion&RecommendationsHardware/SoftwareDimensioningSegmentSize+60%ofpacketsare+500BytesRecommendation:MaximumLLClengthoptimisation(Max.Improvement~1%)HardwareDimensioningTrafficDistribution,BusyHour@2pmwith1/3aboveminimumConclusion&RecommendationsFeaturePriorizationSessionDurationSessionsaremainlyveryshortRecommendation:OverlappingTBFandRapidInterleavingareveryimportantSessionIntervalTypicallybelow5seconds.ShowsbenefitofSuper-CoattailRecommendation:SCTParametersaresetwellConclusion&RecommendationsFeaturePriorizationApplicationUsage97%ofthetrafficisTCPUDPtrafficisWAP(isconnectionoriented)andDNSRecommendation:LLCAcknowledgementnotnecessary案例：引入新应用(PushtoTalk)CaseStudyFive-Introductionofnewapplication&understandingitsimpactCustomTelemetryapplicationwasmodelledandsimulatedbyGSMS8GPRSTimeslotswereconsideredpercell250,500&750usersinacellweresimulated250Users500Users750UsersNewChallengeswithPTToverGPRSPTTisthefirstalways-onapplicationtobedeployedoverGPRSNetworkPDPcontextisalwaysonandTBFissetupduringconversations–couldbeacapacityconstraintRequirementsonGPRSNetworkMinimallatencyOptimalbandwidthtosupportPTTtrafficCellReselection&RoutingAreaUpdatedelaysshouldbeminimalUplink/downlinkTBFreleasedurationshouldbeoptimal,toenablePTTconversationtousethesameTBFandavoidsettingupanewTBF,forevery‘push’OptimiseSGSNReadyTimerValueInSTANDBYnodatareceptionispossible.TogetbacktoREADYstate,themobileneedstobepaged,thusincreasingtheset-uptime,forthefirstpush.Benefit:ByincreasingtheReadyTimerontheSGSN,themobilesstaylongerinREADYstate.Drawback:IncreasedreadytimerincreasesthenumberGMMsignalingmessagesandcauseshigherbatteryconsumption.STANDBYIDLEREADYGPRSattachGPRSdetachREADYtimerexpiryorforcedtoSTANDBYPDUtrans-missionOptimiseduseofPBCCH/PCCHfeaturePBCCH/PCCCHprovidesmoresignallingcapabilitiesforbothGSMvoiceandGPRSservicebymovingallGPRSsignallingontothePBCCH/PCCCH.ItreducesinternalsignallingtrafficovertheGSLandRSLsincetherearefewerrequests,assignmentsorpagingmessagesoverGSLandRSL.ThisisduetoGPRSsignallingtrafficshallbecarriedovertheTRAUGDSOptimiseduseofPBCCH/PCCHfeatureBenefitsforPTTIncreasedsignallingcapacitytosupportPTTtypeapplicationSeparatedsignallingforGPRS(PTT)andGSMFasterpagingShorterTBFset-uptimesinheavilyloadednetworks(cells)comparedtothesituationwithoutthePBCCH/CCCHfeature.DrawbacksByreservingatimeslotforaPBCCH/PCCCH,capacitywillbelostforGPRS,thoughitshouldbenotedthataPBCCH/PCCCHtimeslotmaystillbeusedforuserdata(onlyPPCH&PAGCHi.e.downlink,nouplink).IfacellisRACboundary,theresourceonthecellwillbeusedforlocationareaupdatesandsoleavelessresourceforuserdata.NotallexistingGPRSMSssupportPBCCHandtheseMSswontbeabletoattachwhenPBCCHispresentinthecellOptimisingDRXtimerThegprs_drx_timer_maxparametercontrolshowlongthemobilestationstaysinnon-DRXmode,whenithasleftthepackettransfermodeandentersthepacketidlemode.Themaximumvalueofthisparameteris3seconds.BenefitforPTTDecreasingset-uptimesforconsecutive“pushes”,whendonewithin3seconds(orwhateverdurationisset)afterlastdatatransmission/reception.DrawbackOptimizingthenon-DRXtimerisimportanttoincreaseMSbatterylifetimeOnePhaseaccessOnePhaseaccessisrecommendedforGPRSNetworktoreduceTBFset-uptimes,thisworkswellforlightlyloadednetwork.Foracongestednetwork,thismayresultintoRSLcongestion.ToavoidusingtheRSLtonotifythePCUofanaccess,theBTSforcesthemobiletoperformatwo-phaseaccess,withthemobilesendingaPRRonthereservedblock.SincethisreservedblockisonaPDCH,thissignallingiscarriedoutonGPRSTRAUandnottheRSL,therebyreducingRSLload.(ThisrequiresthatuplinkblocksarereservedforonlyPACKETRESOURCEREQUESTS)ByreservingblocksforPRRsandforcingtwophasepacketaccessRSLcongestionisreduced,butthetradeoffisthattwophasepacketaccessresultsinamuchslowerTBFestablishmenttimethanbothonephaseandenhancedonephase.PTTPerformanceEnhancementGPRS网络无线优化原则及案例分析GPRS网络结构无线环境核心网SGSNGi频率计划UmConf天线PCUPIXDNSCom-HubGGSNGnGb网管系统数据链路DHCPNTPOMC-ROMC-GNTOMC-GUnixISSGPRS优化所涉及的网络单元GrFRAME协议族路由协议隧道协议FRLLC/MACTCP/IPGsAirIPCircuit载干比吞吐量误码率延迟传输速率NetworkCMCC主要统计重传率CS比率无线参数时钟

终端应用测试路测数据GPRS优化所涉及的改善指标GPRS的优化方式SnifferHP37907接口测试MDTT手段工具路测CTPSwordFishgNPMTcpdump/Tcptrace终端用户测试性能测试IP包监测数据库审计应用层传输层MSPCUBTSBSCCorporate

VPNWAP

GatewayRadiusDHCPDNSFRBillingLIANSGSNGGSNISSIOSGPRS测试环境MaxisIntranetAntennaInternetGGSN100bTHub"Radius-Server"(Solaris)BenchmarkServer(Solaris)RSM1RSM2CommHubGn-Desktop(Win2000)SGSNHP37907PCUVLAN20GnLaptop-Testexecution(Win2000)Laptop-MDTT(Win2000)VLAN10GGSN-GiVLAN30WANBorderRouterGi-Trace:Tool:tcpdumpBmServer-Logs/Trace:Tool:tcpdump

log-filesGn-Trace:Tool:windumpSGSN-Logs:Tool:log-filesGb-Trace:Tool:HP37907PCU-Logs:Tool:PCU-LogsMS-Logs/Trace:Tool:log-file

CommView

Perf.MonitorMDTT-Logs:Tool:MDTTMobileOMC-RCallTrace-Logs:Tool:IOSCTP设备链路测试链路GPRS

无线优化原则—数据分析E2E性能—通过DT/CQT对终端性能分析，包括：Attachsuccessrate;PDPactivesuccessrate;RAUsuccessrate;CSusage;Cellreselect;Networklatency,throughput,bandwith;BLERGPRS

无线优化原则—数据分析统计方向系统容量分析

PCU/PRPload分析；

热点地区GPRS时隙分配；

Switch时隙对GSM影响；上下行吞吐量分析；BLER分析；拥塞分析；各信令层丢包分析；GPRS

无线优化原则—数据分析硬件告警分析PCU，RXCDR，BSC，BTS，TCU告警分析；传输信令告警；天馈线系统；

GPRS

无线优化原则—数据分析数据库分析功率控制分析；小区重选分析；邻区拓扑结构分析；GPRS接入及信令Timer;GPRS

无线优化原则—数据分析无线环境分析上下行质量分布；上下行电平分布；C/I分布；Pathbalance分布；TA分布；GPRS

无线优化原则—问题总结系统问题总结系统硬件告警及问题频率干扰问题覆盖问题数据库问题信令配合系统资源问题GPRS

无线优化原则—问题处理无线环境优化系统C/I、BLER分析频率调整覆盖/天线调整相邻小区优化GPRS

无线优化原则—问题处理系统容量与负荷分析PDCH拥塞跟踪及处理；GBL信令链路稳定性跟踪及处理；GBL带宽及配置优化；GSL负荷分析；热点区域负荷分析；核心网容量规模，区域UDBI、USPU负荷趋势

GPRS

无线优化原则—问题处理数据库参数调整针对小区重选，C1/C2调整GPRS接入定时器计数器调整；GPRS释放定时器调整；Gb及小区GPRS容量调整；功率控制调整，功率控制窗及快慢；GPRS

无线优化原则—问题处理GPRS功能问题排查GPRS附着过程GPRS激活过程；路由区域更新过程；异地漫游；P-TMSI重新分配；认证与加密过程；识别过程；寻呼过程；GPRS

无线优化重点GPRS无线网络调整，包含无线资源管理的配置以避免拥塞、载干比优化、在不同编码方式下(CS1,CS2)的覆盖优化等.

1根据GPRS数据用户的实际分布，调整无线接入网的参数和GPRS小区切换调整。PCU数据库参数、Gb接口配置参数的优化。解决系统瓶颈，保证GPRS的提供，重点区域的数据业务质量应作为目前优化的重点.

2

3

4GPRS网络无线优化实例优化实施实例—统计项使用GBLPlanning/TrafficVolumeCalculationGBL_<UL/DL>_DATA_THRPUT_HIST

everye.g.5secondsthePCUgeneratesasnapshot.ThecounteroftheappropriateBINisincremented.However,atthemomenttheBINsexistonlyPCUinternallyandonlytheMean/MaxarereportedtotheOMC.TheMean/Maxvaluesarecomputedbasedonthesamplesgenerated.

Usedto(assoonasBINsareavailableontheOMC–in16.50.f)seethenormaldistributionwithinthelaststatisticinterval(e.g.30min).IncasetheGbtrafficisveryburstytheMeanmightbemisleadingandshownmuchlowervaluesasperiodswithverylowtrafficareconsideredaswell.However,theGbmustcopewiththemeaninbusyperiodsandnotwiththeaverageacrossthe30min.peggingperiod.GBL_<UL/DL>_DATA_THRPUT

thehistorystatsgetonlyasampleseverye.g.5seconds.ThisstatsusesthesampleswithinthisintervalaswellandcalculatestheMean/Max/Minvalues.

Usedto

getaclearpictureoftherealMean/MaxthroughputontheGblink优化实施实例—统计项使用PlanningGPRS_32k_channels_switched,GPRS_channels_switched

eachswitchcosts<800msthatthetimeslotisnotavailableforCSanPS.However,iftheQoS(CSSR,DCR)isn‘treducedletthealgorithmwork.Ifreducedtrytoincreasegprs_reconfig_thresh_idle_tch.Thisdoesn‘treducetheswitchingnumberbutitgivesthesoftwaremoretimetoreact(doesreducePDCHbandwidth).Gbl_Paging_Reqs

howmanypercentofallthepaginggoesthroughtheGPRSnetwork.HowfrequentisthemobilepagedifitisinGPRSactivetransfermode(rightnetworkmodeneedstobeset)MS_Class_<x>_<x>_Req

whatisthecapabilityofthemobilesinthefield.HowmanyPDCHshouldbeequippedtoofferthemax.PerformancetotheuserIMM_Assign_Cause whatforistheAGCHbandwidthused优化实施实例—统计项使用PlanningGPRS_access_per_AGCH,_PCH,_RACH

requestingGPRSaccessonAGCHGPRS_pch_agch_q_length

areadditionalchannelsrequiredGPRS_dynet_failures,_res_reqs,_swi_reqs

dynamicallocationofGDSresources(featurenotusedbyMaxis)优化实施实例—统计项使用MobilityActivityCell_Flush_Reqs

indicatorforoutgoingcellreselections.

Whatfor?Microcellandmovingmobilesarehandedin?Gprs_cell_reselection_attampt,Gprs_cell_reselection_fail onlyforNC2available.SameuseasinGSMthehandoverstatistic.Gprs_ms_neighbor_strong

networkcontrolledreselectionsduetostrongerneighborandnotduetocongestionCoding_scheme_change

calculatethenumberofCSchangesperTBF(otherstats).AnidealTBFhasonly2CSchanges(CS2=>CS3=>CS4).However,itdoesn‘tmeanthatahighCSchangeratepointstowardsaproblem.The/amobilecouldmoveacrossareaswithdifferentRFconditionsandthenCSchangesarerequired.优化实施实例—统计项使用CongestionStatusonCell/PRP<UL/DL>_PDTCH_Q_Length

congestionseverity<UL/DL>_PDTCH_Congestion

howlongisthecellcongested<UL/DL>_PDTCH_Seizure

countsthenumberofTBFs

whatfor?Isoneuser/sessioncausingtheRLCtraffic<UL/DL>_Busy_PDTCH

howmanyPDCHareoccupiedforhowlong(BINsfornumberofPDCHsnotavailableincurrentOMC-R–butinPCU.SameasGBL_<UL/DL>_DATA_THRPUT_HISTstatistic)PRPLoad

howloadedisthePRP优化实施实例—统计项使用AvailabilityCell/PCUAvailable_PDTCH

areallequippedcellsavailable?No_PDTCH_Availmin/max/meanTimeGbl_Unavailable

timenotavailableGPRS_32k_<UL/DL>_Not_AvailGbl_Link_Ins

timeGblinservice优化实施实例—统计项使用AccessibilityChannel_Req_Recvs.Channel_Req_Rej

channel_req_rejispegedfore.g.IA-Rejectincaseof1PhaseaccessandPR-Reject.

MotorolaBSSdoesn‘tsendIA-Reject(forGPRS)asworkaroundimplementationassomemobilesaren‘tcompliantanddon‘tinterpretthemessageproperly.Thus,IA-Rejectwasremovedfromthecode.

TocomputetherejectrateinGSR6theChannel_Req_Sucstatisticwasaddedtocounttherealrejectrate(received–granted)优化实施实例—统计项使用PerformanceTheperformancedoesn‘tonlydependingonRFQuality(thismustbecheckedwithGSMstats–ifTCHsareequippedonthatcell-).Lowcellutilisationcouldbethecase,notproperlyworkingmobiles,mobilityorsomemore)CS12_on_32k_chan

isitworthtokeepthatcarrier32kenabled?Air_<UL/DL>_Data_Blks

computethetransferedbytes,RLCdatathroughput,etc.Air_<UL/DL>_Control_Blks

howefficientarethechannelsused,i.e.percentofcontrolmessages.

TocorrecttheRLCbandwidthandmakeisclosertouserbandwidth优化实施实例—统计项使用Gbl_Flow_Ctrl_Sent

flowcontrolmessagesaresentbeforethePCUbuffergetsflooded.Thiscouldbebecausetheairleakrateisreducedforthatmobile/cellorthedataratereceivedfromtheSGSNisincreased.

Whatfor?

Extremelyhighnumbers(againstthehistoryasusual)couldindicateamobileiskickedbackonRLCbandwidth,PDTCHbandwidthreduced,etc.

IDLE_PDTCH_Intf_Band0

samplestakenonIDLEframeforinterferenceband0(samebandsasinGSM)优化实施实例—Attach信令流程Figure22.Authenticate1.GPRSAttachVLR3.UpdateLocationtoSGSNandVLR4.AttachAcceptHLR1234567890MSBSCSGSNMSC优化实施实例—Attach信令流程Gs优化实施实例—PDP激活信令流程Figure61.ActivatePDPContext1234567890MSBSCSGSNGGSNVLRMSCggsna.b.c.d.bj.zjmnc000.mcc460.gprscmnetsgsnsgsnDNSRootDNSDNSAPN:cmnetcmnet.mnc000.mcc460.gprsggsne.f.g.hAPN:cmnetcmnet.mnc000.mcc460.gprs优化实施实例—PDP激活信令流程GGSNpathtoPDNggsn.bjMnc000.mcc460.gprsbj-DNS1Zj-DNS2cmnetsgsnDNS1RootDNS.zjtestggsna.b.c.dsgsnDNS2APN:test.zjTest.zj.Mnc000.mcc460.gprsTest.zja.b.c.d优化实施实例—PDP激活信令流程GGSNpathtoPDN优化实施实例—PDP激活信令流程Figure62.CreatePDPContext1234567890MSBSCSGSNGGSNVLRMSCDHCP–getsIPaddressBSCPCUMS1MS2Loopback154Loopback2WapGateway72MSAPN12CMNETCMWAPVirtualTemplate4GGSNGGSN44DHCPProxy

4

4MsIPaddress

SGSNDHCPServerCMnetInetnetLoopback354Corp3Corp4

优化实施实例—PDP激活信令流程优化实施实例—PDP激活信令流程Figure63.ActivatePDPContextAccept1234567890MSBSCSGSNGGSNVLRMSC

SGSNGGSNHLRDNSPDNGrGnGiGbBSSActivePDPContextRequestAPNResolutiontoDNSServerAPNResolutionResult-GGSNVTAddressCreatePDPContextRequestAssignIPtoMSCreatePDPContextResponseActivatePDPContextAccept优化实施实例—PDP激活信令流程优化实施实例—GPRS数据传输流程TLLI+NSAPITID优化实施实例—PING性能分析PING的信令解析

上行TBF的建立（channelrequest到立即支配IA）上行传送PCU处理，ULForward到SGSNSGSN处理，及对方服务器的响应下行传送MS，OS，COMPUTER的处理及系统接入（RACHACCESS）优化实施实例—PING性能分析GPRS系统中PING流程优化实施实例—PING性能分析Motorola系统PING分析

测试手机MotorolaT720测试环境按照CMCC的测试规范，完成间隔8秒，payload为500bytes的共10次Ping。

GSN版本：

BSS版本：1670.ClientOS：WindowsXPServeOS：Linux优化实施实例—PING性能分析PING测试时长各个阶段时间分布优化实施实例—PING性能分析各个阶段标准偏差各个阶段时间占用比例分析

根据波动性和耗时比例确定重点问题

使没步的时间达到最短应用时长

确定影响时长的相关因素优化实施实例—PING性能分析改善Ping建议RACH接入保证

繁忙小区RACH碰撞后下次接入时延—

tx_integer

IA保证AGCH资源—bs_ag_blks_res

PDCH资源分配

•DL/UL_PD_Con_TOT确认小区产生拥塞时长

•SDCCH拥塞情况

•IA中LU比重

•Totalcall无线环境及系统资源

Rxlev；RxQual；C/I；Gprs_cell_reselection；Gb；PRPLoading。。。优化实例测试手机Attach问题在使用测试手机+MDTT测试中出现Attach异常的信令流程—手机在发出AttachRequest约1秒后MDTT显示手机收到DetachAccept消息，而后大约6秒钟后才收到IdentifyRequest消息，造成Attach时长为8秒

从L3层信令流程中看出：DetachAccept消息的接收时间为14:04:20.406IdentifyRequest消息的接收时间为14:04:26.715优化实施实例—终端问题举例测试手机Attach问题所有手机接收到的下行L3层的信令消息必然对应着一组时间与其相对应的RLC/MAC层数据包在RLC/MAC层消息中我们可以找到和IdentifyRequest时间相对应的DownlinkDataBlock，但确无法找到和DetachAccept相对应的DownlinkDataBlock。由此得出结论：此问题为手机内部错误造成未响应系统第一次

IdentifyRequest消息，而响应了第二次IdentifyRequest消息，造成了Attach时长长达8秒。优化实施实例—终端问题举例OT96PING问题采用集团公司新的间隔8秒PING测试方法，OT96的丢包率为10%—15%。远远高于其他手机的相同测试结果，及OT96在原有的连续PING的结果。PING失败原因手机未发ChannelRequest消息，PING超时手机晚发送ChannelRequest消息，造成PING超时优化实施实例—终端问题举例成功PING的手机信令流程手机未发送ChannelRequest消息的信令流程手机迟8秒发送ChannelRequest消息的信令流程GPRSQoSFundamentalsQoSintroduction随着数据业务发展,不断增长的互联网应用及其它宽带需求业务(如视频和音乐)需求的增长,增加带宽和购买更快的设备开销很大而效率低.单一的QoS级别是无法解决不同业务的QoS需求的矛盾,它将影响用户的感受度和多种业务的推广,尤其是那些基于IP的视频和话音业务.

QoS提供质量是可预知的,并且是可控的服务.QoS是分段承载的端到端的服务.QoS端到端的分层体系结构QoS的实施一定是端到端的,参考下图,我们可以了解到QoS的这种端到端的分层的体系结构.

GPRSQoSversionGPRSQoS的版本

GPRSR97/98QoSGPRSR99QoS

GPRSR97/98QoS属性

服务优先等级可靠性等级时延等级数据吞吐量等级（平均数据吞吐量和峰值数据吞吐量）

GPRSQoSversionGPRSR99QoS属性

Trafficclass('conversational','streaming','interactive','background')

Maximumbitrate(kbps)Guaranteedbitrate(kbps)Deliveryorder(y/n)MaximumSDUsize(octets)SDUformatinformation(bits)SDUerrorratioResidualbiterrorratioDeliveryoferroneousSDUs(y/n/-)Transferdelay(ms)TraffichandlingpriorityAllocation/RetentionPriority不同版本QoS的映射在现网中,MS,SGSN,GGSN和HLR可能支持不同的QoS版本(R97/98或R99),为了互通,在3GPP规范23.107中定义了不同版本QoS的映射原则.MOTOROLAGSN3.0基于GPRSR99QoS并支持不同QoS版本之间的映射.可能出现的不同版本的QoS互连情景如下表所示:

情景编号

MSSGSNGGSNIR97/98R99

R97/98IIR97/98R99

R99

IIIR99

R99

R97/98IVR99

R99

R99

R99QoS规范是基于DiffServ框架,支持会话级,流级,交互级,和背景级等QoS级别以适应不同的业务性质及不同的流量需求.DiffServFundamentalsDifferentiatedServicesArchitecturalModelThedifferentiatedservicesarchitectureisbasedonasimplemodelwheretrafficenteringanetworkisclassifiedandpossiblyconditionedattheboundariesofthenetwork,andassignedtodifferentbehavioraggregates.EachbehavioraggregateisidentifiedbyasingleDScodepoint.Withinthecoreofthenetwork,packetsareforwardedaccordingtotheper-hopbehaviorassociatedwiththeDScodepoint.

DifferentiatedServicesDomainADSdomainisacontiguoussetofDSnodeswhichoperatewithacommonserviceprovisioningpolicyandsetofPHBgroupsimplementedoneachnode.ADSdomainhasawell-definedboundaryconsistingofDSboundarynodeswhichclassifyandpossiblyconditioningresstraffictoensurethatpacketswhichtransitthedomainareappropriatelymarkedtoselectaPHBfromoneofthePHBgroupssupportedwithinthedomain.NodeswithintheDSdomainselecttheforwardingbehaviorforpacketsbasedontheirDScodepoint,mappingthatvaluetooneofthesupportedPHBsusingeithertherecommendedcodepoint->PHBmappingoralocallycustomizedmapping[DSFIELD].DSBoundaryNodesandInteriorNodesBothDSboundarynodesandinteriornodesmustbeabletoapplytheappropriatePHBtopacketsbasedontheDScodepoint.Inaddition,DSboundarynodesmayberequiredtoperformtrafficconditioningfunctionsasdefinedbyatrafficconditioningagreement(TCA)betweentheirDSdomainandthepeeringdomainwhichtheyconnectto.DSIngressNodeandEgressNodeDSboundarynodesactbothasaDSingressnodeandasaDSegressnodefordifferentdirectionsoftraffic.ADSingressnodeisresponsibleforensuringthatthetrafficenteringtheDSdomainconformstoanyTCAbetweenitandtheotherdomaintowhichtheingressnodeisconnected.ADSegressnodemayperformtrafficconditioningfunctionsontrafficforwardedtoadirectlyconnectedpeeringdomain,dependingonthedetailsoftheTCAbetweenthetwodomains.DifferentiatedServicesRegionAdifferentiatedservicesregion(DSRegion)isasetofoneormorecontiguousDSdomains.DSregionsarecapableofsupportingdifferentiatedservicesalongpathswhichspanthedomainswithintheregion.TrafficClassificationandConditioningDifferentiatedservicesareextendedacrossaDSdomainboundarybyestablishingaSLAbetweenanu