Traffic analysis on the WIDE backbone link: From transport level to end user activity

We conduct the analysis of the traffic measured on the backbone link between the WIDE network that connects a number of leading universities in Japan and the upstream internet service provider. We study its statistical properties such as the packet sizes and time distributions between packets. To reveal the laws governing the end user activity and its impact on the network traffic dynamics, we next extracted inter-packet time and packet sizes for TCP, UDP and ICMP protocols. Additionally, long-term correlation properties of the traffic have been analyzed and its scaling characteristics were estimated. We suggest a statistical model which could be used to describe and simulate the backbone link traffic that is in substantial agreement with the empirical observations both in terms of its statistical and correlation properties.