Scaling Behaviors of Traffic in Computer Communication Networks

In this paper, we present an empirical study on the scaling behaviors of traffic in computer communication networks. It is shown that the data shipped by the web server have a symmetry heavy-tailed distribution, with the corresponding cumulative distributions of both negative and positive tails can be well fitted by a Levy regime or an extendedly asymptotic power-law form. We investigate the long-range dependence of the absolute value of both packet-changes and byte-changes by using the detrended fluctuation analysis, and find a crossover at approximately 5 minutes (300sime10^2.5 seconds) in both the two cases. The corresponding scaling exponents are 0.66/0.98 and 0.63/0.97 for the measure of packets and bytes, respectively