Study On New Methods Of Traffic Matrix Analysis

Internet technology is one of the fastest growing technology in twenty-first Century. It has been widely used in our life, and made great contribution to social and economic development. However, with the development of Internet technology, many new network applications have emerged in recent years. Though they provide people lots of convenience and entertainment, they add great burden to the network providers. Meanwhile, different access ways provided by a large number of heterogeneous networks makes Internet difficult to be controlled. So how to analyze and manage the network effectively is very important.Traffic matrix, as an important input parameter of network traffic engineering, has been widely studied by researchers at home and abroad. Study on traffic matrix includes two directions, traffic matrix analysis and traffic matrix estimation. This thesis shows a new analysis method to study and analyze traffic matrix, called diffusion wavelets-based method, which is put forward in recent years. At the same time, based on this method, this thesis develops a method to detect and analyze traffic matrix abnormality. The main contents of this paper are divided into the following three aspects:1) Operator selection. Diffusion wavelets-based method is a multi-resolution analysis method. Diffusion wavelets can decompose the traffic matrix into the approximation matrix and the detail matrix on different scales by using a diffusion wavelet operator. These coefficient matrices are the key data for anomaly detection experiments. Through the experimental analysis, different diffusion wavelet operators produce different wavelet coefficient matrices, the differences bring different analysis results and anomaly detection performances. So the first work of this paper is the analysis and comparison of two commonly used diffusion wavelet operators, Random-Walk operator and I-L operator. Secondly, network anomaly is studied by using an appropriate operator-based diffusion wavelet analysis.2) Anomaly detection. Anomaly detection experiment is introduced after diffusion wavelet operator comparison experiment. Anomaly detection experiment is divided into the anomaly detection algorithm design and the selection of experimental data. Finally, this thesis gives the result of anomaly detection.3) Anomaly positioning. At the end, this thesis found the potential rules between diffusion wavelet coefficient matrix and the original traffic matrix through the experiment and statistics. By this rule, we can infer position of the anomaly node in the original traffic matrix from the abnormal changes of the coefficient matrix. As an application of this rule, this paper designs an experiment of open circuit detection.Multi-resolution analysis based on diffusion wavelet can be used to analyze original traffic matrix through coefficient matrix on an appropriate scale. This not only reduces the complexity of calculation, but also improves the analysis accuracy and efficiency.