Non-stationary and Small-time Scaling Behavior of Internet Traffic

Abstract

Scaling behaviors of traffic influence network modeling, service providing and traffic engineering. So it has been the focus of many researches. In this paper, an extensive wavelet analysis is performed to the TCP flows. The results show that the TCP flows exhibit non-stationary and uncorrelated scaling behaviors at the small-time scales(1? 100ms). To identify the non-stationary and small-time scaling causing factors, the flow composition of the traffic is analyzed along two dimensions-flow byte contribution and flow density. And the results show the dense flow is the non-stationary and uncorrelated scaling causing factor in small-time scales.