Analysis And Research Of Network Traffic Model Based On Fractal Theory

The analysis of network performance and communications network planning and design is mainly based on traffic model. Accurate traffic model is of great significance to the design of high-performance network protocols, business forecasting and network planning, high-performance network equipment and servers, accurate network performance analysis and prediction, congestion management. Traditional telecommunications networks made full use of Poisson model to describe the random properties of network traffic in order to study the number of calls, waiting time, call duration and other parameters of the statistical characteristics in the telephone communications networks and thus design telephone networks, in the case of guaranteeing a certain Quality of Service . Practice has proved that this model using in the traditional telephone network system design and performance evaluation problem is very effective.With the development of communication technologies, communications networks have also rapidly changed. Researchers quoted the Poisson model in early research of data service network traffic. Network traffic described by Poisson model have no aftereffect and smooth characteristics. And also with the rapid development of internet network, there are a variety of new business as well as the proposed business model. The traditional Poisson model is no longer useful for internet. After in-depth study, the researchers introduced some of a little complex random process. Then along with discovery of self-similar characteristics of network, the researchers have conducted many self-similar network traffic models. However, self-similarity is a form of single fractal, it is inevitably difficult to resolve the complexity of network traffic behavior.This topics concern to many fractal wavelet model is based on a multi-fractal model. The only difference from the single fractal is that the model not only the application of the fine scale wavelet analysis, but also the application of the multi-scale analysis characteristics of fractal theory ,thus is a good solution to describe the network traffic long-range dependence and short-related issues at the same time. However, the model was first proposed mainly in order to line up the analysis of simulation, the description of actual traffic random properties are not enough, the model needs to be improved and in-depth studied.This paper mainly works on the following aspects:1. In this paper, we introduced the mathematical definition of some network traffic characteristics, such as self-similar, long-range dependence, multi-fractal and so on. We also analyzed their corresponding physical meaning and compared difference between traditional traffic model and the self-similar traffic model. And then we summarized modeling theory of some mainstream fractal network traffic model following, such as fractal Brownian motion, fractal Gaussian noise, fractal ARIMA process, discrete wavelet model and so on.2. Because of its fine-scale analysis characteristics, Wavelet transform began to be used on multi-scale analysis of network traffic. It combined with fractal theory to provide network traffic in different time-frequency-scale diversification measure. After establishment of multi-fractal wavelet model, There are still some problems and the model needs to be improved. In this paper, we take full account of the various flow characteristics which would be fitted actual statistical distribution, when proposing appropriate structural-scale coefficient and wavelet coefficients. In this way we can select a suitable statistical distribution which is more in line with the actual traffic characteristics. Finally we compared it with existed model in order to analyze differences between them.3. In this paper, we introduced huge challenges that network traffic behavior abnormal detection is currently facing, mainly as a result of DDoS attacks and the proliferation of P2P traffic. On the base of introducing the Distributed Denial of Service attacks and peer to peer technology, we provided a certain idea for detection by corresponding analysis of traffic characteristics..