Research of Network Traffic Model Based on Multifractal Wavelet
|School||University of Electronic Science and Technology|
|Course||Communication and Information System|
|Keywords||Network flow model Self-similarity Wavelet transform Multi-fractal distributed Traffic generation system|
Network traffic model is a basis of network design and analysis, and the preciseflow model can accurately portray the actual network traffic characteristics, networkperformance analysis, network parameter design, network congestion control is of greatsignificance.Self-similar network traffic model can describe network traffic characteristicsaccurately, and has a wide range of applications. But self-similar model is the adderstructure, there is a serious negative flow problems. In addition, self-similar networktraffic model can only describe the scale structure of the linear form, which is difficultto accurately describe the complex scaling properties of network traffic. Themulti-fractal wavelet network traffic model in wavelet analysis as a tool to achievemulti-scale characteristics of network traffic analysis, it could also describe long in thetime scale of the network traffic as well as small time scales multi-fractal. This paperfocuses on the independent wavelet model (IWM) and multi-fractal wavelet model(MWM) cannot ensure that the signal is non-negative coefficients structure, selectdistribution and the actual traffic distribution independent of the characteristicslimitations, reasonable selection coefficient structure distribution, building accuratelycharacterize network traffic characteristic behavior of network traffic model designed tosimulate real network traffic characteristics of the nature of network traffic generatingsystem. The main research contents and innovations are as follows:Firstly, the network traffic characteristics of multi-scale are analyzed. Actualnetwork traffic in different time scales within the multi-fractal characteristics. And thenselect a single parameter self-similar traffic model is not comprehensive enough, needto select the multi-fractal wavelet model. Compared and analyzed the IWM and MWMmodeling idea and its limitations.Secondly, Proposed a multi-fractal mixed wavelet network traffic model based onthe Pareto distribution. the model taking into account the IWM good long descriptionability and short-term burst MWM ability to describe, by setting the randommultiplication factor limiting characteristic coefficient of non-negative to ensure that reconstruction flow is positive, selected in line with the actual flow is heavy-taileddistribution characteristics of the Pareto distribution, the more accurate reconstructionof traffic signals, improved network traffic model is verified by experiments inperformance is greatly improved, and in line with the actual flow of network trafficcharacteristics.Thirdly, a distributed traffic system based on the multi-fractal wavelet model isproposed. The system uses the multi-fractal wavelet flow model reconstruction traffic astraffic data sources, control terminal sender to multiple to allocate traffic generationtasks, distributed by the sender synthetic test traffic. Distribution of traffic generationsystem to meet the multi-user concurrent access to the actual network environment,distributed to generate test traffic, accurately simulate different network applicationservices.Through simulation, P_OWM preferably enhanced in the multi-fractal spectrum,the marginal distribution, the correlation function and other performance indicators, isbetter than the MWM and closer to the actual flow. The probability density distributionof reconstruction flow is a heavy-tailed distribution, in line with the characteristics ofthe actual traffic distribution.