Estimating effective bandwidths from traffic data

The effective bandwidth of a traffic source is a measure of its resource requirements that is most relevant for quality of service provisioning in broadband networks. It has to be estimated from observed source output, but direct estimators that have been proposed for this purpose are biased for autocorrelated traffic. By explicitly computing the bias for autoregressive and Markov sources, we devise a bias correction scheme that does not require knowledge of the model parameters. Its use enables us to obtain accurate estimates of effective bandwidths from fairly short data records.