QoS Monitoring

QoS Monitoring
Dr.MohitBansal Profile Pic
Dr.MohitBansal,Canada,Teacher
Published Date:26-10-2017
Your Website URL(Optional)
Comment
QoEandQoSMonitoring Withthegrowthofmobileservices, ithasbecomeveryimportant foran operator to measuretheQoSandQoEofitsnetworkandcustomersaccurately.Themeasurements canbeusedtoanalyseproblematiccasesandimprovetheperformance(optimise)ofthe networkorserviceseffectivelyandcost-efficiently.Thatwillinturnhelpmaintainboth customerloyaltyandcompetitiveedge. Inthischapter,whattheend-userperceivesismainlyanalysedintermsoftheintegrity oftheservice,whichconcernsthroughput,delay(anddelayvariationorjitter)anddata loss of a bearer service during communication. Service accessibility, which relates to availability,security,activation,access,coverage,blockingandsetuptimeofabearer service, and retainability, which in general characterises connection losses, are only partially covered in this book. Such performance metrics may be straightforwardly derivedbysignallingmessageflowchartsavailableinthestandardsorbymonitoring directlythemessagesexchangedbetweennetworkdomainsthroughthecorresponding interfaces. MeasuringQoSinthenetworkisofvitalimportanceforperformancemanagement and optimisation. The ability to measure QoE will help the operator to gather the contributionofnetworkperformancetotheoveralllevelofcustomersatisfaction. Inthischapter,anin-depthdiscussionisprovidedaboutdifferentaspectsofQoSand QoEassessments,includingmeasures,frameworks,methodsandtoolsusedtoachieve QoSandQoEtargets. 9.1 QoE and QoS assurance concept QoEandQoSmonitoringincellularnetworksconsistsofcollecting/processingservice and network performance statistics, usage data and QoS-related faults. In order to obtainend-to-endqualityofservicemonitoring,networkelements(NEs),theelement managementlayer(EML)andnetworkmanagementlayer(NML)mustallbeinvolvedwiththeQoEandQoSmonitoringprocess.Alarmandperformancecollection(counters and/or gauges) is performed at the network element layer and alarm/performance aggregation,reportgenerationandanalysisisdoneatelementmanagementandnetwork managementlayers. TheQoEandQoSmonitoringprocessconsistsofthefollowingfunctions: . ManageQoSfaultconditionsreceivedfromnetworkelements. . Retrieve QoE and QoS performance data from network elements, which includes mobileterminals. . Collectandprocessusagedata. . GenerateQoEandQoSreports. . UndertaketrendanalysisofkeyQoEandQoSparameters. . Audit/AnalysecollectedQoEandQoSparametersagainstexpectedvalues. 9.1.1 Conceptualarchitecture ThearchitectureofaQoE/QoSmonitoringsystemisshowninFigure9.1.Thearchi- tecturalcomponentsshowninthefigurearedescribedinthefollowingsections. Theconceptualarchitectureformeasurementresults,aggregation,transferandpre- sentationmaybedesignedasdepictedinFigure9.2.Theelementmanagementlayerand networkmanagementlayerprovidethreedistinctfunctions: Itf-N Figure 9.1 QoEandQoSmonitoringconceptualarchitecture1.Use Use Planning and Benchmarking Care of MVNO and Helpdesk Marketing and case case optimization travellers SLA sales sells Coverage vs. Power -45.00 10.00 -50.00 6.79 -55.00 5.83 5.00 -60.00 0.00 -65.00 -67.33 -65.25 -68.12 -71.01 -69.37 -70.00 -73.17 -5. 00 -75.00 -10.00 -80.00 -85.00 -15.00 -90.00 -93.68 -95.00 -97.39 -98.57 -21.13 -20.00 -100.00 -22.50 -103.98 -25.88 -25.00 -105.00 -110.00 -30.00 Display V D V D V D V D V D V D 12307 12922 12341 11274 10047 Display Ce llI Ds RSSI RSCP UeTXpwr Visualize Logic Logic Logic Logic Logic Logic Analyse Analyse Post- processing Report Report Data Admin. and data storage 2 Figure 9.2 Conceptual architecture for data post-processing, visualisation (V) and drill-down (D) in particular locations/positions or in the raw performancemetricsusedintheanalysiscriteria. RSSI and RSCP (dBm) UeTxPower (dBm)318 QoSandQoEManagementinUMTSCellularSystems . Administration and data storage (reportlayer)in acommon database: this is where measurementsandconfigurationfilesarecollected. . Post-processing (analyse layer): this is the logic used for intelligent data analyses accordingtoparticularusecases.Severalanalysismethodologiesmaybeimplemented dependingonthedesiredlevelofaccuracyandresolution(seeChapter10). . Visualise(displaylayer):thisiswheretheperformanceresultsattainedfromintelligent dataanalysesarepresentedandvisualised(denotedbyVinFigure9.2)accordingto networkandserviceadministratorneeds.Theresultsmaybefurtherdisplayedusinga ‘drill-down’approach(denotedbyDinFigure9.2).The‘zoom-in’ofmonitoreddata makes itpossibletoinvestigate,forexample,whathappenedinawkward positions (street,area,etc.)orlocations(cell,RNC,etc.)whenamissingserviceorpoorQoS performancewasdetected.Thedrill-downapproachmayalsobeusedforidentifying whycertaincriteriaindataanalyseswereunsatisfactory,bydisplayingthecombined (raw)metricsseparately. IntheusecasesdepictedinFigure9.2,theresultsofQoEandQoSperformancedata analysesareusedfornetwork(service)planningandoptimisation;benchmarking(e.g., between different terminals and/or network/service providers); caring of corporate or otherparticularcustomers;monitoringofservicelevelagreements;improvinghelpdesk efficiency;sales,marketingandproductmanagement;mobilevirtualnetworkoperator (MVNO)cases,etc. 9.1.1.1 Network element Thenetworkelement(includingthemobileterminal,UE)isresponsibleforcollecting performance measures, usage data and generating alarms. It supports the following functions1: .Collectperformancedataaccordingtohowmeasurementshavebeendefinedand returnresultstotheelementmanagementlayer. . Collectusagedataandforwardthedatatomediation. . Performfaultdetection,generationofalarms,clearingofalarms,alarmforwarding and filtering, storage and retrieval of alarms in/from the NE, fault recovery and configurationofalarms. 9.1.1.2 Element management layer Theelementmanagementlayer(EML)isresponsibleforaggregatingandtransferring collectedQoEandQoSperformancemeasurementsandgeneratedalarms/events.The EMLconsistsofthefollowingfunctions1: . performancemanagement(PM);and . faultmanagement(FM).QoEandQoSMonitoring 319 Performance management ThePMfunctioncollectsmeasurementdata.Measurementtypesandmeasurednetwork resourcesforQoEandQoSassurancearediscussedinSections9.2and9.3,respectively. TherelatedtoolsaredescribedinSections9.6.1and9.6.2,respectively. Theresource(s)towhichthemeasurementtypesareappliedhavetobespecified;this includesthedefinitionofameasurementrecordingperiod(periodsoftimeatwhichthe NEcollectsmeasurementdata),measurementreportingcriteriaandmeasurementreport fileformat.Themeasurement-relatedinformationtobereportedhastobespecifiedas partofthemeasurement.Thefrequencyatwhichscheduledresultreportsaregenerated hastobedefined. Measurement results can be transferred from the NE to the EM according to the measurementparameters.TheymayalsobestoredlocallyintheNEandcanberetrieved whenrequired.MeasurementresultsmayalsobestoredintheEMforretrievalbythe networkmanager(NM)whenrequired. Fault management ThemainFMfunctionisthemanagementofalarmeventreports(e.g.,mappingofalarm and related state change event reports, real time forwarding of event reports, alarm clearing)andretrievalofalarminformation(e.g.,retrievalofcurrentalarminformation on NM request, logging and retrieval of alarm history information on NM request). 9.1.1.3 Network management layer From the QoE and QoS monitoring perspective, the NML is responsible for the collection and processing of performance, fault and usage data. The NML QoE and QoSmonitoringlayerprovidesthefollowingfunctions1: . Service quality management (SQM): this is responsible for the overall quality of a service as it interacts with other functional areas to access monitored information, process that information to determine quality metrics andinitiate corrective action when the quality level is considered unsatisfactory. Inputs to SQM include both performanceandfaultdata. . CustomerQoSmanagement(CQM):thisincludesmonitoring,managingandreporting the QoS customers receive against what has been promised them in service level agreements and any other service-related documents. Inputs to CQM include data fromSQMandusagedata. An example of a complete service assurance solution for the network management system(NMS)ispresentedinSection9.7. 9.2 QoE monitoring framework AsdiscussedbrieflyinChapter1,therearetwopracticalapproachestomeasuringQoE inmobilenetworks.320 QoSandQoEManagementinUMTSCellularSystems 1. Servicelevelapproachusingstatisticalsamples. 2. NetworkmanagementsystemapproachusingQoSparameters. These two methods are not mutually exclusive. If these two approaches are used to complementeachother,theQoEmeasurementsaremuchmoreaccurateandrealistic fromtheend-userpointofview.Forexample,theoperatorcouldidentifytherootcauses ofserviceapplicationperformancenotbeingmet,orrelateQoEperformancemeasures toothermetricsmeasuredinthenetwork(suchasthroughput)inordertoevaluatethe spectralefficiencyoftheparticularinterfaceatwhichthroughputiscollected. ToolsusedforQoEmonitoringareintroducedinSection9.6.1.Acompletesolution forserviceassuranceispresentedinSection9.7. 9.2.1 Servicelevel approachusing statisticalsamples Keytothefirstapproachisstatisticalsamplingandtakingthemostrelevantandaccurate measurementsaccordingtothatsample.Inthisapproach,mostmeasuredperformance indicatorsareatapplicationlevel,providingtherealend-userperspective. The service level measurement approach relies on a statistical sample of overall networkuserstomeasuretheQoEforallusersinthenetwork.Ifthesizeofthestatistical sample(numberofobservations)iscorrectlyselected,theresultswillgetclosetoachiev- ing100%precision(repeatabilityofmeasures),muchlikethemostreliablepollsbefore elections. Section 9.4 provides some insight into various statistical methods and con- fidenceintervals. Thisprocessinvolves: . Determining the weighting of key service applications.Manyserviceapplicationsare usedbyvariousnetworkusers.Therightservicemixhastobeestablishedandreflect therightkindofservicesmostlyusedinthenetwork.Atthesametime,someapplica- tionsaremorepopularandmorefrequentlyusedthanothers.Also,someservicesare moreimportanttoend-usersthanothers,eveniftheirusagefrequencyisnotveryhigh. Basedonsomenetworkstatisticshistory,marketunderstanding,trendsandoperator preferences, each application in the selected service mix needs to be given aweight percentage.Thetotalweightoftheservicemixneedstobe100. . Identifying and weighting QoE KPIs.Eachapplicationhas uniquekeyperformance indicators(KPIs)thatneedtobeidentified.Variousstudieshavebeenmadeonthis subjectanddifferentstandardisationbodieshavedefinedtheseKPIsintheirownway asgeneralguidelines.TheseguidelinescanbeusedasabasistofindKPIsforeach application.Ontopofthat,theweightofeachKPIhastobedefined,asitvariesfrom application to application. For example, jitter could have almost no impact on perceived performance in the case of file downloading, but it can have significant impactonstreamingandvideosharingtypesofapplications.Thenetworkpointof measurementofeachKPIalsoneedstobedefinedforproperreference. . Devisingaproperstatisticalsample(geographicareas,trafficmix,timeofday,etc.)and collectingQoEKPIs.Thisisthemostimportantstepinthisapproachandiskeytothe accuracyofresults.Thehelpofaprofessionalstatisticianisrequiredtoensureproper sampling. A good sample would consider the proper representation of all kinds ofQoEandQoSMonitoring 321 users,includingvariousterminals,properservicemixandtheirweights,usagepatterns liketimeoftheday,variousgeographicdistributions,etc. . Utilisingmobileagentsinhandsetstomaketheresultsmoreaccurate.Themobileagents inthehandsetscanprovideextrainformationonthisprocess.Thesemobileagentscan be installed in the handsets of selected users from a carefully selected sample. In practice, various strategies can be deployed. For example, incentives can be given tothoseuserswhoagreetoinstallmobileagentsintheirhandsets.Informationfrom these agents needs to be collected on a regular basis for incorporation into other networkmeasurements. . GivinganoverallQoEscore(index)fromKPIvaluesforeachseparateserviceanda service mix.Somekindofspreadsheetorapplicationwouldbeneededtocalculatea finalQoEindexbasedontheinputsprovided.Thetoolneedstobeveryflexibleand adaptabletochangingcircumstances–likenewapplications,newusagepatterns,etc. Regular revision of the above-mentioned steps would ensure accurate and realistic QoEindexdetermination.WeakareasneedtobeimprovedensuringbetterQoE. Thisapproachhasmanybenefits,forexample: . The QoE for any network can be measured without access to operator products/ NMSs,etc. . OperatorscanalsomeasureQoEprovidedbyitscompetitors. . Itisvendor-independentand,toalargeextent,bearer-independent. . Inexpensivetoolscanbeusedandtheprocessishighlyscalable. . Goodforbenchmarkingvariousservicesandapplicableglobally. Despitebeingveryuseful,regulardrive-throughtestingisnottheonlyanddefinitelynot themostpracticalwaytoimplementthisapproach.Severalcombinationscanbeapplied, including, for example, the mobile agents in UEs already mentioned above, the data collected from some performance measurement tools which use remote probes and stethoscopes,etc. Frequenttestsmayimprovetheaccuracyofresults. 9.2.2 NetworkmanagementsystemapproachusingQoS parameters IntheNMSapproach,hardQoSperformancemetricsfromvariouspartsofthenetwork aremappedontouser-perceptibleQoEperformancetargets.TheseQoSmeasurements are made using a network management system, collecting KPI figures from network elementsandcomparingthemwiththetargetlevels.Theprocessinvolves: . Identifying the relationship between QoS KPIs and their effect on QoE.Thisisthekey stepinthisapproach.Thisareahasbeenunderthespotlightforsometimeandseveral studieshavelookedintounderstandingtheexactrelationshipbetweensomeQoSKPIs andend-userperceptionofperformance(QoE).Althoughprogressisbeingmade,due to the many practical problems acomprehensive solution is not straightforward to achieve.Addingvariousbitsandpieceswouldstillmakesomekindofrelationship modelthatcanbeusedinthisapproach,thoughaccuracyofQoEdeterminationwith thisapproachalonemightbequestionable.322 QoSandQoEManagementinUMTSCellularSystems . Measuring QoS KPIs in the network.Variousnetworkmanagementsystemsandtheir supplementstakelotsofQoSmeasurementsfromthenetwork.Itwouldbepractically impossibletomeasureallparametersforallusersatalltimes,thusagoodstatistical samplingofusersandserviceswouldberequiredinthismethodaswell. . Rating QoE through measured QoS KPIs using some mapping metrics. Mapping of various QoS KPIs onto user perception of performance will need to be made on a spreadsheet,forexample.TheinputsfromhardQoSperformancemetricswouldbe fedintothatsheettocalculatetheQoEbasedontheidentifiedmapping. Inprinciple,thisapproachcouldbeidealforoperators,thoughitdoesdependonhow wellthemapping reflectsrealend-userexperience.Identifyingtheproperrelationship betweennetworkQoSKPIsanduserexperience(QoE)isanextremelychallengingtask, butisveryunlikelytoprovidethehighestlevelofaccuracy.Thisapproachisverymuch vendor-orientedandlacksflexibilityandscalability. The best and more realistic option would be to use both these methods in a com- plementaryway.Togetherthesetwomethodswillprovidethemostaccuratepictureof userexperienceandgivethenetworkoperatorthepossibilitytoassesstheutilisationof radioandtransportresourcesatagivenQoE(percentageofsatisfiedusers). 9.2.3 QoEmetrics AlthoughQoEissubjectiveinnature,atthesametimeitisveryimportanttohaveit measured. Waiting for end-users to vote with their money might turn out to be very expensive for stakeholders. As such, a strategy has to be devised to measure QoE as realisticallyaspossible. Thetop-downapproachcouldbeusefulinthisregard: . Thekeyistounderstandthefactors(metrics)contributingtouserperception. . Applythatknowledgetodefinetheoperatingrequirements(values). . Deviseamethodologytomeasurethesefactorsconstantly(tools,location,statistical sampling)andimprovethemasandwhenneeded. WhenidentifyingQoEmetrics,wemustfirstaskatabroadlevel,‘Whatdoestheend- userexpectfromanyoralloftheentitiesintheQoEvaluechain?’(seeFigure1.2).There willbeasmanydifferentexpectationsasthereareusers,butmostoftheseexpectations canbegroupedundertwomaincategories: reliabilityand quality. . Reliability,inthiscontext,istheavailability,accessibilityandmaintainabilityofthe content,theservicenetwork,and/ortheend-userdeviceandapplicationsoftware. . Quality,ontheotherhand,referstothequalityofthecontent,thebearerservice,and/ ortheend-userdeviceandapplicationsoftwarefeatures. Havingestablishedthetwodimensionsinwhichend-userswilljudgetheirQoE,wecan nowhavealookattheperformancemetricsthatdescribethesedimensionsfromthe network viewpoint. The metrics are listed in Table 9.1. It is however important to evaluate these KPIs, first, on the basis of the four QoS classes (conversational, inter-QoEandQoSMonitoring 323 Table 9.1 QoEkeyperformanceindicators. Reliability (service quality of accessibility and retainability) QoEKPI Mostimportantmeasures Explanation Service Ratioofterritoryunder Territorycoveredwithinacountry. availability coveragetonotunder Sincetheword‘universal’isan (anywhere) coverage(%) importantcomponentofUMTS,itis quitelikelythatmanyuserswillfind coveragetobeanimportantissueintheir judgementofquality.Operatorsshould plantheirnetworkandalsotheirroaming agreementsinsuchawaythatthey satisfyeverysinglesegmentofusers. Globalcoverageusingroamingpartners canbeincludedlaterifrequired. Service Ratioofrefusedconnections Onceauserknowsthatserviceis accessibility orratioofPDPcontext availableinaparticulararea,itis (anytime) failedtoestablishinfirst importantthattheserviceisalwaysup attempt(%) andaccessibleatalltimes.‘Pleasetry again’messagesarefrustratingwhenone needstodownloadanimportantemail. Serviceaccess Averagecallorsessionsetup Ifserviceisaccessibleatanygiventime,it time time(s) isimportantthatthecallsetuporsession (servicesetup establishmenttimetoaccesstheserviceis time) nottoolong.Longwaitingtimescan frustratetheuser. Continuityof Serviceinterruptionratio Nothingcanbemoreexasperatingthanto serviceconnection (%) becutofffromtheservicewhileaccessing (service (using)it. retainability) Quality (service quality of integrity) QoEKPI Mostimportantmeasures Explanation Qualityofsession Serviceapplicationlayer Thisisameasureofthenumberofpackets packetlossratio(%) lostoutofathousandpacketssentduring asession.Userswillalwaysbemore toleranttopacketlossinconversational andstreamingapplications,thanin backgroundandinteractiveapplications. Fortheseapplications,TCP/IP retransmissionsaremeasuredasan indicationofpacketlossandhencedelay, loweredthroughput,highresponsetime andnetworkcongestion. Bitrate Averagebearerbitrate Thismeasureisespeciallyimportantfor achievedasratioofbitrate conversational,streamingandsome demandedbyapplication interactiveserviceapplicationssuchas (%) interactivegaming.ThePDPcontext allocatedbitratecanchangeduringthe session,asnetworkconditionschangeor324 QoSandQoEManagementinUMTSCellularSystems Table 9.1(cont.) Quality (service quality of integrity) QoEKPI Mostimportantmeasures Explanation asaresultofusermobility.Hence lookingattheaveragebitrateovera sessionisimportant.Ifthebitrate achievedislowerthantheencodedbit rate,theuserwillhaveafrustrating experience. Bitratevariation Bearerstability:bitrate Auserwillalwayspreferastablebit variationaroundnegotiated rate,especiallyforrealtimeandstreaming bitrate(%) applications.Thismeasurecalculatesthe standarddeviationaroundtheaveragebit ratemeasuredabove. Activesession Averagethroughputtowards Itisnormallyassumedthatanend-user throughput mobile(kb/s) whohasahighthroughputtowards his/hermobilewillalwayshaveahigh QoE.Thisassumptionholdstrueformost interactiveandbackgroundclass applicationsandhenceitisimportantto includethismeasureinQoEcalculations. However,aswaspointedoutabove,in certain applications, as long as the average bearerbitrateachievedishigherthanthe bitratedemandedbyanapplication,there willbenoperceptibledifferenceintheQoE experiencedbythetwousers. System Averageresponsetime(s) Instreamingandsomeinteractive responsiveness applications,itisimportanttolook separatelyattheaverageresponsetime. Theaverageresponsetimeisthetimeit takesthefirstpacketsofinformationto arriveafterarequesthasbeensent (providedthePDPcontexthasalready beenestablished).Itdoesnotapplyto certainapplicationssuchasvoiceand videoconferencing. End-to-enddelay Averageend-to-enddelay End-to-enddelayherereferstothetimeit (msors) takes a packet to traverse the network and getfromhosttodestinationor vice versa. Conversational,streamingandcertain interactiveclassapplicationsarevery muchdelay-sensitive.Alongdelaywill leadtoverynegativeend-userexperience. Delayvariation Jitter(%) Evenifdelayisminimised,variationinit couldleadtopoorQoE.Henceitis importanttofactorinthejitterwhichis thestandarddeviationexperiencedinthe delaytimesofthepacketsduringa particularsession.QoEandQoSMonitoring 325 active, streaming and background) and, second, on the basis of the characteristics of someofthemostpopularserviceapplicationsthatfallundereachofthesetrafficclasses. In this section, we mostly focus on 3GPP QoS models; insight into the approaches adopted in other standardisation bodies is given in the sections that follow. More information on service applications and corresponding target performances can be foundinChapter2. Thevalueofeachofthesemetricswouldtranslatetodifferentlevelofimpactonthe actualQoE.Somewillbetotallyirrelevantinonecasewhilebeingthemostimportantin another. It all depends on the type of service application being run by the user. Characteristics are different and as a result the requirements for all applications may notbethesame. Inordertoreducethescopeofperformanceanalysis,itwouldbegoodtounderstand thedifferentserviceapplicationsintermsofthefourtrafficclassesdefinedin3GPP.As mentionedabove,thefourtrafficclassesareconversational,streaming,interactiveand background. Voiceandvideoconferencingcallsarethetwomostpopularserviceapplicationsthat needtheconversationalclasstrafficattributespecifiedin3GPPR99.Bothapplications sharecertaincharacteristicsthatarecommontoservicesthatfallunderthistrafficclass: . Theypreservethetimerelation(variation)betweeninformationentitiesofthestream tominimisedelayvariation. .Aconversationalpatternthatisstringentandlowdelay. . Theyarerelativelyinsensitivetopacketloss. . Guaranteedresourceallocation,noretransmissions. . Realtimetraffic. Audioandvideostreamingarethetwomostpopularserviceapplicationsthatshoulduse streamingclass.Fortheseserviceapplications,wecanstatethat: . Ashortdelayisimportantfortheapplication. .Alowbiterrorrateisoftenrequested. . Importanttominimisedelayvariation. . Guaranteedresourceallocation. . Lightretransmissionsareused. Web and WAP browsing, remote server access (Telnet) and interactive gaming are envisagedasthefourmostpopularapplicationslikelytousetheQoStrafficattributes assignedforinteractiveclass.Theimportantcharacteristicsoftheseserviceapplications are: . Request–responsepatternofuse. . Preservepayloadcontent(i.e.,minimisebiterrorrate). . Transmissiondelaymustbeacceptableforinteractiveuse. . Dynamicresourceallocation. . Retransmissionsareused. . Best-efforttraffic(non-realtime).326 QoSandQoEManagementinUMTSCellularSystems There are many service applications that fall under background class. These include email,messagingandFTP.Themostimportantcharacteristicsofbackgroundclassare: . Someamountofdelayistolerablebytheapplication. . Lowerprioritythaninteractivetraffic. . Responsivenessofinteractiveapplicationsisensured. . Applicationsneedreliabledata,transmission,delayisnotimportant. . Dynamicresourceallocation. . Retransmissionsareused. . Best-efforttraffic(non-realtime). 9.2.3.1 ETSI view ETSIhasputalotofeffortintoexplainingQoSmetricsfromtheend-userpointofview (QoE)in2.Alldefinedqualityofserviceparametersandtheircomputationsarebased on field measurements. That indicates that the measurements were made from the customerpointofview(fullend-to-endperspective,takingintoaccounttheneedsof testing).Bynomeansisitacompletespecification,butitstillcontainsenoughinforma- tiononthissubjecttowarrantreading.TheQoEmetricsidentifiedandgroupedinthe specificationsareinlinewiththedescriptionmadeatthebeginningofthissection.Figure 9.3illustratesthis. Network Layer 1 access Network non- Circuit Packet accessibility switched switched (NNA) Service access Service integrityService Layer 2 retainability Telephony Layer 3 SMS CSD PSD ... ... Service accessibility Service accessibility Service accessibility Service accessibility SMS MO telephony (SA-T) CSD (SA-CSD) PSD (SA-PSD) (SA-SMS MO) ... ... Access delay PSD Access delay SMS Access delay CSD Setup time telephony (AD-PSD) (AD-SMS) (AD-CSD) (ST-T) Speech quality End-to-end delivery Data quality (DQ) Data quality (DQ) (SpQ) time SMS (DT-SMS) Call completion Call completion rate Session completion rate CSD telephony (CCR- rate PSD (CCR-CSD) CS) (SeCR-PSD) Figure 9.3 ETSIQoSparametersfromend-userperspective2.QoEandQoSMonitoring 327 9.2.3.2 ANSI view ANSIT1.522-2000specifiestheclassesofQoSsufficienttosupportbusinessmultimedia conferencingonIPnetworks,definedasbeingequivalenttolegacyconferencesystem performance.Inthisstandard,twobaselineconferencingsystemsaredefined3: . Tier1:desktopPCsystems. . Tier2:groupconferenceroomsystems. . The specification defines the perceptible performance level and the acceptable per- formance level. For some aspects (e.g., bit rate and loss) there are different levels specifiedforTiers1and2.Theperformanceparametersofinterestaresummarisedin Tables9.2and9.3. Table 9.2 ANSI-definedqualityofserviceparameters–multimedia(MM)userinterface3. Parameter Qualitycriteria Communication Speed Accuracy Availabilityandreliability function Connection Setuptime Mis-directed Accessibilityratio(amongmedia) establishment Transfertime Connectionfailure Userinformation Delay(spontaneity) Mediaquality Droppedconnection Delayvariation Media Contention synchronisation resolution Connection Take-downtime Releasefailure release Table 9.3 ANSI-definedqualityofserviceparameters–end-to-endinterface3. Parameter Qualitycriteria Communication Speed Accuracy Availabilityand function reliability Connection Setuptime Mis-directed Accessibility establishment Transfertime Connectionfailure Userinformation Delay(networklatency) Losttransport Dropped (packet)transfer Delayvariation(within PacketRate(combines connection asinglemediastream IPpacketDefects,such (IPavailability) andbetweenstreams) Error&LostPackets) Informationbitrate (committedbitrate anddeliveredbitrate) Connection Take-downtime Releasefailure release328 QoSandQoEManagementinUMTSCellularSystems Error- Conversational Voice/video Streaming audio Fax tolerant voice and video messaging and video Transactions Messaging, Command/control Error- (e.g. e-commerce, Downloads Background (e.g. Telnet, intolerant WWW browsing, (e.g. FTP, (e.g. Usenet) interactive games) email access) still image) Interactive Responsive Timely Non-critical (delay 1 s) (delay 2 s) (delay 10 s) (delay 10 s) Figure 9.4 ITU-Tmodelforuser-centricQoScategories4. 9.2.3.3 ITU-T view ITU-TRecommendationG.1010definesamodelformultimediaQoScategoriesfroman end-user viewpoint 4. By considering user expectations for a range of multimedia applications,eightdistinctcategoriesareidentified,basedontolerancetoinformation lossanddelay.ItisintendedthatthesecategoriesformthebasisfordefiningrealisticQoS classes for underlying transport networks and associated QoS control mechanisms. Figure9.4describesthismodel. InaninformativeannextoITU-TRecommendationG.10104,indicationsofsuitable performancetargetsforaudio,videoanddataapplicationsaregiven.Thesetargetsare reproducedinTables9.4and9.5. 9.3 QoS monitoring framework Thissectionpresentsmeansandmethodsofmeasuringthequalityofdistinctservices acrossUMTScellularnetworks.Theapproachisbasedonmappingserviceapplications withdifferentperformancerequirementsontodistinctPDPcontexts(QoSprofiles).By meansofasubsetofbearerserviceattributes(bitrates,prioritiesandQoSclasses),itis possibletoformulatemetricstomeasureseparatelytheperformanceofserviceswithin differentnetworkdomains,inuplinkanddownlinkdirections,withoutanyvisibilityof the content carried by upper layer protocols. Besides this, we present performance indicators to monitor the utilisation of interfaces between network domains and per- formancedeteriorationduetohardwarelimitationsinnetworkelements.Theproposed measures collected by access and core network elements provide essential inputs to ensurequalitycompliancetoservicelayermanagementcommitmentsandoptimisation solutions.Table 9.4 ITU-Tperformancetargetsforaudioandvideoapplications4. Medium Service Degreeof Typicaldata Keyperformanceparametersandtargetvalue application symmetry rates One-waydelay Delayvariation Informatonloss Other Audio Conversationalvoice Two-way 4–64kb/s 150mspreferred 1ms 3%PLR – (e.g.,telephony) 400mslimit Audio Voicemessaging Primarily 4–32kb/s 1sforplayback 1ms 3%PLR – one-way 2sforrecord Audio High-qualitystreaming Primarily 16–128kb/s 10s 1ms 1%PLR – audio one-way Video Videophone Two-way 16–384kb/s 150mspreferred 1%PLR Lip-synch: 400mslimit 80ms Video Broadcast One-way 16–384kb/s 10s 1%PLR Assumesadequateechocontrol. Exactvaluesdependonspecificcodec,butassumesuseofapacketlossconcealmentalgorithmtominimiseeffectofpacketloss. Qualityisverydependentoncodectypeandbit-rate. Thesevaluesaretobeconsideredaslong-termtargetvalueswhichmaynotbemetbycurrenttechnology. PLR¼Packetlossratio.Table 9.5 ITU-Tperformancetargetsfordataapplications4. Medium Serviceapplication Degreeof Typicalamount Keyperformanceparametersandtargetvalues symmetry ofdata One-waydelay Delayvariation Informationloss Data Webbrowsing–HTML Primarilyone-way 10kB Preferred 2s/page N.A. Zero Acceptable 4s/page Data Bulkdatatransfer/retrieval Primarilyone-way 10kB–10MB Preferred 15s N.A. Zero Acceptable 60s Data Transactionservices–high Two-way 10kB Preferred 2s N.A. Zero priority(e.g.,e-commerce, Acceptable 4s ATM) Data Command/control Two-way 1kB 250ms N.A. Zero Data Stillimage One-way 100kB Preferred 15s N.A. Zero Acceptable 60s Data Interactivegames Two-way 1kB 200ms N.A. Zero Data Telnet Two-way 1kB 200ms N.A. Zero (asymmetric) Data Email(serveraccess) Primarilyone-way 10kB Preferred 2s N.A. Zero Acceptable 5s Data Email(server-to-server Primarilyone-way 10kB Canbeserveral N.A. Zero transfer) minutes 6 Data Fax(‘realtime’) Primarilyone-way 10kB 30s/page N.A. 10 BER 6 Data Fax(store&forward) Primarilyone-way 10kB Canbeserveral N.A. 10 BER minutes Data Low-prioritytransactions Primarilyone-way 10kB 30s N.A. Zero Data Usenet Primarilyone-way Canbe1MBor Canbeserveral N.A. Zero more minutes N.A.¼Notavailable.QoEandQoSMonitoring 331 An introduction to tools used for QoS monitoring is given in Section9.6.2andan exampleofacompletesolutionforserviceassuranceisreportedinSection9.7. 9.3.1 Performancemonitoringbasedonbearerservice attributes WhenaPDPcontextisestablished,theattributesofthenegotiatedQoSprofileforthat particular bearer service are available in the UE/MS, RNC/BSC, SGSN and related GGSN.Toassessserviceapplicationscarriedatdifferentnetworkelements(i.e.,RNC/ BSC, SGSN and GGSN), bearer service performance needs to be ascertained from a subsetofattributesofthesubscribedQoSprofile–forexample,trafficclass(TC),traffic handlingpriority(THP)forinteractiveclass,allocationretentionpriority(ARP)andbit rates(maximumandguaranteed)–whichunambiguouslyrelatetotheofferedservice application(one-to-onemappingbetweensubsetsofPDPcontextattributesandupper layer protocols). (This concept was presented in Chapter 8 for consistent and differ- entiatedQoSprovisioning.)Forinter-workingbetweendifferentreleases,thefollowing mapping(provisioning)rulesbetweenR97/98andlater3GPPreleasesapply5(seealso Section3.2.2.11andChapter8): . DelayClass1correspondstointeractiveTHP1. . DelayClass2correspondstointeractiveTHP2. . DelayClass3correspondstointeractiveTHP3. . DelayClass4correspondstobackgroundTC. Furthermore,forconsistenttraffictreatmentthroughthemobilenetwork,weneedtoset ARP¼THP¼Precedence class¼Delay class. Background class (Delay Class 4) is mapped by the 2G SGSN onto ARP¼Precedence Class 3. Hence, for (E)GPRS R97/98, without taking into account the bit rate attribute, four combinations can be defined, although in the BSS only two combinations (i.e., Precedence Class 3 and background class, which corresponds to Delay Class 4) undergo the same treatment (only three precedence class settings are available); whereas for 3G, again without consideringthebitrateattribute,tenpossiblecombinationsofattributes(PSconversa- tional ARP1–3, PS streaming ARP1–3, interactive THP1–3 and background) can be definedforlegacyandQoS-awareterminals.Thepracticalcombination,definitionand mappingofservicesontotheseparticularsubsetsofQoSattributesisfortheoperatorto manage. Thismeansthatcountersinnetworkelementsshouldbeclassifiedwithagranularity that allows the NMS to make statistics based on these particular combinations of attributes. In Figure 9.5, we illustrate an example of how different counters collected by 3G/2G SGSN and RNC/BSC can be classified to support PDP context based monitoringofthroughputbythemanagementlayer.Theclassificationofmeasurements bytheRNC/BSCand3G/2GSGSNincludesattributes–suchastrafficclassandtraffic handlingpriority(inthecaseofinteractiveclass),orprecedenceclassanddelayclass– thatenabletheNMStofilteroutindicatorsandcomputethroughputperbearerservice. In the following sections, we present some metrics for differentiated QoS analyses. Fromthesemeasuredmetrics,takingintoaccountone-to-onemappingofapplications ontodistinctQoSprofiles,itispossiblefortheoperatortodrawconclusionsonthe332 QoSandQoEManagementinUMTSCellularSystems RNS/BSS throughput per 3G/2G SGSN throughput per subset of QoS attributes subset of QoS attributes Collect classified counters from NEs and Subsets of QoS compute throughput using that particular attributes (‘pipe’) subset of QoS attributes for • Definition consistent performance monitoring Management layer RNC/BSC 3G/2G SGSN Delivered RLC data blocks Delivered BSSGP/GTP data blocks Classification of counters: Classification of counters: • 3G: traffic class, THP (1, 2, 3) • 3G: traffic class, THP (1, 2, 3) • 2G: precedence class (1, 2, 3) • 2G: delay class (1, 2, 3, 4) Network layer Figure 9.5 ExampleofhowclassifiedcountersatdifferentNEscanberetrievedbythemanage- mentlayertocomputethethroughputofaparticularbearerservice(‘prioritypipe’)consistently. performance of each provisioned service application. However, if more services (or media components in the case of 3GPP R6 or later releases of IMS) are carried by thesamePDPcontext,orplacedbehindthesameaccesspointname(APN)forlegacy terminals (i.e., mapped onto the same subset of QoS attributes on which metrics are classified),QoSmonitoringofservicesusingthismethodisnotpossible.Inthiscase,only the overall performance of the corresponding traffic aggregate (multiplexed IP flows) wouldbeassessed,andtheintegrity,retainabilityandaccessibilityofdifferentservices cannolongerbedistinguishedfromeachother. 9.3.2 QoSmonitoringinBSS ThemonitoringprocessisimplementedinallnetworkelementsoftheBSS.Bearerservice performance needs to be assessed at each layer of the user plane protocol stack (see Figure 4.1). For each protocol, different performance metrics can be collected. For example, at the RLC level it is possible to measure RLC throughput and BLER (RLC SDU error ratio), if AM RLC is used. Also, at the BSSGP level, the BSC is abletomeasureessentialmetricsforassessingsystemperformance(e.g.,BSSGPbuffer utilisation,BSSGPthroughput,ratioofdiscardedBSSGPPDUspercellorMS).Other keyperformanceindicatorsthatreflectthequalityexperiencedcanbederivedfromBSC/ PCUstatistics.Forexample,thepagingsuccessratiorelatestoaccessibilityoftheservice, whilePDTCHblockingdealswithavailabilityandtheTBFestablishmentsuccessrate givesanideaoftheintegrityoftheservice. Troubleshooting activities lead to a second tier of performance indicators. This is needed for identifying the network conditions that deteriorate performance – forQoEandQoSMonitoring 333 example,PSterritoryutilisation,MSmultislotrequested/allocatedratio,MCS(orCS) usageratio. In order to correctly interpret the values of measured metrics, it is important to understandtheobject/levelaggregationofthestatisticsgivenbythenetworkmanage- mentsystem.Moreover,fortroubleshootingpurposes,valuesattheRLClevel(within agreed quality criteria)mayindicategoodRF performance, butnot necessarily satis- factory QoE: upper layers may be underperforming, for instance. Hence, complete analysis of performance at different protocol layers must be carried out. In this case, measuringLLCpacketdeliverysuccessandtheIPpacketlossratiowouldbeessential inputstoQoEperformancemonitoring. In the following sections, some examples of performance indicators are presented. Othermeasuresthatenableanalysisofcurrentnetworkresourcesaswellastheacces- sibility, availability and integrity of the session are also possible. Each vendor (or operator) has its specific means and methods of implementing a monitoring process, whichincludescountersatBSSlevel,formulasforcomputingperformanceindicators andtoolsformeasurementresults,aggregation,transferandpresentationintheNMS. 9.3.2.1 Classification of counters in the BSS IntheBSS,countersfordifferentiatedperformancemonitoringneedtobecollectedin bothuplinkanddownlinkdirections–separatelyforGPRSandEDGEnetworks–and classifiedbasedon: . IMSI. . RLCtransmissionmode(acknowledged,unacknowledged). . EDGEmodulationandcodingschemes(MCS-1toMCS-9). . GPRScodingschemes(CS-1toCS-4). . Precedenceclassorallocationretentionpriority(1–3). . Cellidentifier(cellID)orBSSvirtualconnectionidentifier(BVCI); . AttributesofBSSQoSprofile6: e peakbitrate,codedasthevaluepartinthebucketleakrate–downlinkanduplink; e typeofBSSGPSDU(signallingordata)–downlink; e typeofLLCframe(ACK,SACKornot)–downlink; e precedenceclassusedatradioaccess(1–4)–uplink. Thismakeitpossibleforthenetworkadministratortocollectmeasurementsforcon- sistentserviceperformancemonitoringacross(E)GPRSnetworkdomains,asexplained in Section 9.3.1, or for gathering performance metrics according to specific needs. In practice, only a subset of the above attributes may be required. For example, in the followingsectionsonlytheprecedenceclassattribute,denotedby p,isused. 9.3.2.2 Integrity monitoring Inthissection,wepresentsomemetricstoassesstheintegrityofaserviceapplication carriedonaparticularprecedenceclass(orARP,asdiscussedinSection9.3.1)inthe BSS.334 QoSandQoEManagementinUMTSCellularSystems BLERmonitoring InAMRLC,BLERcanbederivedfromtheRLCSDUerrorratio,whichindicatesthe fractionofincorrectlyreceivedRLCblocksoverthemeasurementperiodforacertain connection.Thus,thefollowingratioforEGPRSBLERcomputationcanbeused: MCS-9 X p N CorrectRxRLCblocks;i i¼1 BLER ¼1 ð9:1Þ p MCS-9 X p N TotalTxRLCblocks;i i¼1 where N isthenumberofRLCblockscorrectlyreceivedbytheMSor CorrectlyRxRLCblocks PCU (BSC), N is the total number of RLC blocks sent for each of the TotalTxRLCblocks modulationandcodingschemesemployed(MSC-1toMSC-9)and pistheprecedence classsetintheQoSprofile.ThesameformulacanbeusedforcalculatingtheSDUerror ratio in the GPRS BSS, in this case MSC-1 to MSC-9 is replaced by CS-1 to CS-4. Throughput computation p;k Let B bethenumberofdeliveredRLCblocks(withoutretransmissions)ofTBF iof i p durationd ,wherepistheprecedenceclass,andkisamodulationand/orcodingscheme i thathas13possiblevalues(CS-1toCS-4,MCS-1toMCS-9).Totalcorrectlydelivered p p bitsb andtherelateddurationD inacellofthe(E)GPRSnetworkcanbecalculatedas: MCS-9 N N X X X p;k p p p b ¼ r B ; D ¼ d ð9:2Þ k i i k¼CS-1 i¼1 i¼1 wheretheradioblocksizeinbitswithoutRLCheaderr dependsontheCSorMCSkof k thatparticularTBFi(seeTables4.1and4.2),andNisthetotalnumberofTBFscollected by the BSC in that particular cell during the measurement period S. The average throughputperuserinthecellinquestion,from(9.2),istherefore: p b p t ¼ ð9:3Þ p D p Ultimately,foreachprecedenceclass p,meancellthroughput t canbederivedfrom cell (9.3)bysummingupallcorrectlydeliveredGPRSandEGDEbitsinthecell,overthe monitoredperiod S,anddividingtheattainedvalueby S;thatis: p b p t ¼ ð9:4Þ cell S Anexperimentalvalidationoftheabovethroughputanalysiswaspresentedin7. 9.3.2.3 Accessibility monitoring ThissectionpresentstwoexamplesofserviceaccessibilitymeasuresfortheBSS.

Advise: Why You Wasting Money in Costly SEO Tools, Use World's Best Free SEO Tool Ubersuggest.