Multi-scaling behaviors in internet traffic

Scaling refers to the absence of a particular characteristic time scale controlling a stochastic process. The literature on network traffic widely agrees with a mono-scaling Gaussian self-similar traffic model at the large time scales. In this paper, several studies are performed on the Internet traffic. The results show that differs from traditional works, the traces exhibit a multi-scaling, non-Gaussian Poisson behavior at the sub-second time scales. So, we claim that a non-stationary Poisson model needs to be taken into account when modeling network traffic at the sub-second time scales.