On the Autocorrelation Structure of TCP Traffic TITLE2
Daniel R. Figueiredo, Bingqiu Liu, Vishal Misra, Don Towsley · 2000
Considerable attention has recently focussed on the characteristics of network traffic and related issues. In particular, the observation that data traffic in networks exhibits long range dependence or self-similarity has led to a myriad of research efforts in this area. In this paper we show that the TCP protocol can generate traffic with correlation structures over a finite range of timescales that are similar to characteristics observed in Internet traffic. This is independent of application-level characteristics and takes effect under a wide range of network conditions. We point out and analyze two of the mechanisms inside TCP that contribute to this: the exponential timeout back-off and the congestion avoidance. We support our claim using Markovian models that can be numerically analyzed. The models are validated against simulation results and measurements obtained in the Internet.