There is experimentally observed self-similar, or long range dependent, behaviour in local area network (LAN ) traffic [1] and VBR video traffic [2].
The cause of self-similarity [6] in traffic statistics in real networks is an open question. We approach this problem in the belief that many traffic flow control protocols are intrinsically capable of generating chaotic behaviour, which we believe to be self-similar. We have observed [4] chaos [3] in computer experiments [5] on a simple model of a communications network, and here we report self-similarity in the system dynamics.
The model consists of a two-dimensional network of routing cells for data transfer, governed by deterministic rules, whose dynamics are complex [4], [5]. This network is a feasible means for transferring data, and arguably representative of the more complicated real systems which are used.
The dynamics of the system display self-similarity, measured by the Hurst parameter [6] which when calculated for the the average packet age time series, is close to unity (which value would imply the maximum possible degree of self-similarity.)
The irregular flows observed in real computer networks may be a consequence of the dynamics of the protocols rather than irregularities in the load or service demand. This means that the statistics of the models normally adopted are based upon invalid assumptions.