Research On Network Traffic Modeling, Analysis And Web Traffic Control

With the rapid development of the Internet, modeling and control of network traffic become more and more important. This dissertation focus on modeling and analysis of video conference traffic, video movie traffic and network traffic, and control of web traffic. The main results and contributions of this dissertation are as follows:1. For the self-similar and monofractal video traffic, a rayleigh-wavelet model (RWM) is proposed. In scaling space and wavelet spaces, RWM uses different methods to model the scaling and wavelet coefficients, so the proposed model can match the marginal distribution and the long-range dependence of the video traffic. Moreover, under the multiscale queue (MSQ) framework, we give a theoretic formula of the queueing analysis.2. For the multifractal video conference traffic, a gamma-beta based multifractal model is proposed. By using a beta-distributed multiplier, this model can match the gamma distribution of the video conference at almost all time-scales.3. For the multifractal network traffic in WAN, a new multifractal wavelet model is proposed. At all time scales, this model shapes the scaling coefficients to match the marginal distribution of the source traffic. Moreover, due to the scene change nature of video movies, we propose an effective way to determine the coarsest scaling coefficients of the multifractal wavelet model for the video movie traffic.All the above models pay more attention to the marginal distributions of the network traffic at different time scales, thus they are suitable for queueing analysis for network link buffers with different critical time scales, and they can get more accurate results than existing network traffic models.4. The control of web traffic is solved. By using a MIMO model and the wavelet based optimal experiment design methodology, we identify and control the web server. Through this web server control system, we can partially eliminate the self-similar nature of the web traffic, reduce the cell lost rate of the network link buffers, improve the bandwidth utilization and the web server's performance effectively.