无线区域网路中自我相似交通流量之

無線區域網路中自我相似交通流量之成因與效能評估Theoriginandperformanceimpactofself-similartrafficforwirelesslocalareanetworks報告者：林文祺指導教授：柯開維博士OutlineBackgroundofSelf-SimilarityPropertiesofWLANTrafficEstimationofSelf-SimilarTrafficTheOriginofSelf-SimilarityinWLANImpactofSelf-SimilartoCSMA/CAperformanceImpactofSelf-SimilartoCSMA/CAperformancewithRTS/CTSBackgroundofSelf-Similarity(1/8)Self-SimilarityandFractalBackgroundofSelf-Similarity(2/8)StatisticsofSelf-SimilarityDefinitionofSelf-SimilarStochasticProcess:,0HExatExtaa2HVarxatVarxta2(,),xHRatastsaxRH:Hurstparameterorself-similarityparameterBackgroundofSelf-Similarity(3/8)Self-SimilarityofStatisticsDefinitionofSelf-SimilarStochasticSequence:()(1)1,1,2,3,...kmmkiikmmxxkm3323133kkkkxxxx(),12mVarxVarxHmEx.mxxRkRkBackgroundofSelf-Similarity(4/8)PropertiesofSelf-SimilarityLongrangedependenceSlowlydecayingvarianceHeavy-taileddistribution,01Ckkask,11()Pr[]~,0FxXxxx,當()mVarxVarxmBackgroundofSelf-Similarity(5/8)Self-SimilarTrafficBackgroundofSelf-Similarity(6/8)X(t)isaParetodistributionrandomprocesswithshapeparameterαandlocationparameterk.1()fxkkx1,;0kFxxkx,11EXk222,221k32HParetoDistribution:BackgroundofSelf-Similarity(7/8)Variance-timePlot(),1,2,3,...mVarxVarxmm～loglogmdVarxdm()log[()]~log[()]log()mVarxVarxm1(/2)HR/SPlotjX=inflowduringyearj,1jNMN=constantyearlyoutflowjL=Reservoirlevelatendofyearj,1jN11NjjMNXN1jjkkLXjMN11maxminjjjNjNRNLL211NjjSXMNNloglogRNaHNSNBackgroundofSelf-Similarity(8/8)PropertiesofWLANTraffic(1/2)WLANtrafficTimeUnit=1SecTimeUnit=0.1SecTimeUnit=0.01SecBasic:1μSAggregation:1,0.1,0.01SecEnvironment:7NBPoissontrafficPropertiesofRealNetwork(2/2)TimeUnit=0.01SecTimeUnit=0.1SecTimeUnit=1SecEstimationofSelf-SimilarTraffic(1/2)•PacketsSequenceonWLANEstimationofSelf-SimilarTraffic(2/2)•VariancePlot&R/SPlotSingleSourcewithoutCSMA/CATheOriginofSelf-SimilarTraffic(1/3)TheOriginofSelf-SimilarTraffic(2/3)•VariancePlot&R/SPlotTheOriginofSelf-SimilarTraffic(3/3)•VariancePlot&R/SPlotforWLANbasedonsinglePoissonTraffic.(Simulated)ImpactofSelf-SimilartoCSMA/CAperformance(1/7)MaximumthroughputTheinfluenceofnodesonSelf-SimilarTrafficandPoissonTrafficTheinfluenceofpacketlengthonSelf-SimilarTrafficandPoissonTrafficImpactofSelf-SimilartoCSMA/CAperformance(2/7)012345678010203040NodeThroughput（Mbits）SS1000SS2000SS250P1000P2000P250405060708090100010203040NodeUtilization（%）SS1000SS2000SS250P1000P2000P250Maximumthroughput00.0010.0020.0030.0040.0050.0060.007010203040NodeAveragepacketdelay（sec）SS1000SS2000SS250P1000P2000P250ImpactofSelf-SimilartoCSMA/CAperformance(3/7)Maximumthroughput0501001502002503003504004500510152025303540NodesNumberofcollisionsSS1000SS2000SS250P1000P2000P25000.511.522.533.544.5020406080Load（%）Throughput（Mbits）SS,N=5SS,N=10SS,N=20P,N=5P,N=10P,N=2001020304050607080020406080Load（%）Utilization（%）SS,N=5SS,N=10SS,N=20P,N=5P,N=10P,N=20ImpactofSelf-SimilartoCSMA/CAperformance(4/7)TheinfluenceofnodesonSelf-SimilarTrafficandPoissonTraffic050100150200250020406080Load（%）NumberofcollisionsSS,N=5SS,N=10SS,N=20P,N=5P,N=10P,N=2000.00050.0010.00150.0020.0025020406080Load（%）Averagepacketdelay（sec）SS,N=5SS,N=10SS,N=20P,N=5P,N=10P,N=20ImpactofSelf-SimilartoCSMA/CAperformance(5/7)TheinfluenceofnodesonSelf-SimilarTrafficandPoissonTraffic40506070809010020406080100Load（%）Utillization（%）SS1000SS2000SS250P1000P2000P250ImpactofSelf-SimilartoCSMA/CAperformance(6/7)TheinfluenceofpacketlengthonSelf-SimilarTrafficandPoissonTraffic22.533.544.555.566.5720406080100Load(%)Throughput(Mbits)SS1000SS2000SS250P1000P2000P25005010015020025020406080100Load（%）NumberofcollisionsSS1000SS2000SS250P1000P2000P25000.00020.00040.00060.00080.0010.001220406080100Load（%）Averagepacketdelay（sec）SS1000SS2000SS250P1000P2000P250ImpactofSelf-SimilartoCSMA/CAperformance(7/7)•TheinfluenceofpacketlengthonSelf-SimilarTrafficandPoissionTrafficMaximumthroughputTheinfluenceofnodesonSelf-SimilarTrafficandPoissonTrafficTheinfluenceofpacketlengthonSelf-SimilarTrafficandPoissonTrafficImpactofSelf-SimilartoCSMA/CAperformancewithRTS/CTS(1/4)11.522.533.544.555.560510152025303540NodeThroughput（Mbits）SS1000SS2000SS250P1000P2000P25050556065707580850510152025303540NodeUtilization（%）SS1000SS2000SS250P1000P2000P250ImpactofSelf-SimilartoCSMA/CAperformancewithRTS/CTS(2/4)Maximumthroughput11.522.533.5020406080Load（%）Throughput（Mbits）SS,N=5SS,N=10SS,N=20P,N=5P,N=10P,N=2001020304050607080020406080Load（%）Utilization（%）SS,N=5SS,N=10SS,N=20P,N=5P,N=10P,N=20ImpactofSelf-SimilartoCSMA/CAperformancewithRTS/CTS(3/4)TheinfluenceofnodesonSelf-SimilarTrafficandPoissonTraffic11.522.533.544.555.5620406080100Load（%）Throughput（Mbits）SS1000SS2000SS250P1000P2000P250354045505560657075808520406080100Load（%）Utilization（%）SS1000SS2000SS250P1000P2000P250ImpactofSelf-SimilartoCSMA/CAperformancewithRTS/CTS(4/4)TheinfluenceofpacketlengthonSelf-Simil