Starting from the analysis of traffic data, accurately collected on an Ethernet network, the authors discovered a characteristic nobody expected: they demonstrate that Ethernet LAN traffic is statistically self-similar. None of the commonly used traffic models is able to capture this behavior, which has great implications for the design, control, and analysis of high-speed, cell-based networks.

The paper contains a description of the available traffic measurements; a theoretical presentation of self-similarity; the statistical analysis of the collected data for self-similarity; and a discussion on the significance of self-similarity for traffic engineering of B-ISDN (broadband ISDN) environments.

All the sections of the paper contribute to the demonstration of the traffic self-similarity characteristic. Details on the network environment at Bellcore are given. References to papers describing self-similarity are included. The statistical analysis is presented in detail. The impact of the self-similar nature of traffic on the engineering of high-speed networks is also highlighted.

Network engineers will find this paper interesting and useful.