A parallel method for generating self-similar network traffic

After a long period of measurement on network traffic, the results have shown that the packet traffic is associated with a long-range correlation that can be described by using statistically self-similar processes. One difficulty in the research of modern network is how to generate large amount of network traffics efficiently and effectively. This paper presents a parallel method for generating self-similar network traffic method to solve this problem. We then discuss the performance in using such generating method for simulating network traffic under the environment of Nankai Star* (A Cluster System). The results represent that long traces can be generated in much less time. We also conduct experiments to show that our method can generate self-similar traffic for specified Hurst parameters with high accuracy.