Identifying patterns in Internet traffic

The increasing role of information technology has resulted in high demand for Internet services. Consequently, Internet traffic modeling has become an essential field of study. Due to the inadequacy of existing traffic models, new traffic models are needed to allow proper network design and dimensioning. In this paper the patterns of Internet traffic type,,; are discussed. Based on the real network trace collected from the Access Network, Pearson System Analysis is done to identify a univariate distributions for each traffic type. But chi-square test of fit failed. The experimental results using the statistical pattern matching technique, Vector Quantization gives models fitting each Internet service traffic. These models fit with 0% error The training is done with five days of data, testing with 30 days of data. These models can be used in performance analysis or simulation of traffic types, and to identify the misuse of a traffic type.