Mathematical analysis of network traffic

The level of complexity in computer networks is rising and the mathematics needed to model the network traffic behaviour is not an exact science. This paper aims to bridge the gap between mathematics and engineering by illustrating some of the problems that exist with conventional traffic modeling, and show how to obtain informative network statistics via mathematical tools such as the Hurst (1951) parameter and the autocorrelation function. We show how aggregated traffic behaves over various time scales and focus on certain protocols to observe their impact on the network at various ingress/egress points on our university network. Furthermore, we present the many analytical tools that are useful in characterising these systems.