Evolution of the subspace method to ATM traffic analysis-cell loss approximation for actual ATM traffic

This paper describes evolution of the subspace method in signal processing to a cell loss approximation of actual asynchronous transfer mode (ATM) traffic. This approximation can be applied to the actual traffic where the burst lengths are general distributions (burst intervals are exponential distributions) without fall back. Furthermore, it permits real time processing without the use of high speed and often expensive processors. This objective cannot be achieved by conventional operations research (OR) approaches. The comparisons with simulations show that the results of this approximation are within 95% confidence when burst lengths ate Γ-distributions and constant distributions. We concluded that this approximation can be applied to actual ATM traffic analysis