On the applicability and utility of Gaussian models for broadband traffic

Abstract

Gaussian traffic models for high speed communication networks have received an increasing amount of attention. There are some puzzling and counter-intuitive features to deal with, such as negative traffic. However, on the plus side, the central limit theorem can be invoked to argue that traffic will become closer to Gaussian as more and more of it is aggregated together. The Gaussian model has a number of interesting implications for connection admission control and network architecture, which are discussed in this paper. Finally, there is an important distinction between continuous time and discrete time models which we need to make. It appears that the continuous time Gaussian models currently under study are significantly different from the discrete time models discussed in this paper.