Long range mutual information

Network traffic modeling generally views traffic as a superposition of flows that creates a timeseries of volume counts (e.g. of bytes or packets). What is omitted from this view of traffic is the contents of packets. Packet contents (e.g. header fields) contain considerable information that can be useful in many applications such as change and anomaly detection, and router performance evaluation. The goal of this paper is to draw attention to the problem of modeling traffic with respect to the contents of packets. In this regard, we identify a new phenomenon: long range mutual information (LRMI), which means that the dependence of the contents of a pair of packets decays as a power of the lag between them. We demonstrate that although LRMI is hard to measure, and hard to model using the mathematical tools at hand, its effects are easy to identify in real traffic, and it may have a considerable impact on a number of applications. We believe that work in modeling this phenomenon will o...

• Type: Journal
• Year: 2008
• Where: SIGMETRICS
• Authors: Nahur Fonseca, Mark Crovella, Kavé Salamatian
• Keywords: Hardware, Network Traffic, Router Performance Evaluation, SIGMETRICS 2008, Volume Counts