Estimation of token bucket parameters for aggregated VoIP sources

The paper presents an analytical study aimed to establish a dimensioning procedure for the token bucket algorithm, used as a meter in a Differentiated Services network architecture, when a stochastic model for the multiplexed traffic is available. In the work, we propose an equivalent queueing system method to 'on line' estimate the linear bounded arrival processes (LBAP) parameters when a non-zero probability of non-conforming packets is accepted. Then, we validate our approach considering an aggregation of fluidic On-Off processes with exponentially distributed sojourn time in each state, used to model telephone sources with voice activity detection (VAD). To test the goodness of our analytical results, we employ discrete event simulations, which have highlighted the accuracy of the proposed dimensioning procedure in a voice over IP (VoIP) scenario. Moreover, we investigate the multiplexing gain and the effect of different parameters on the LBAP characterization of the VoIP traffic, taking into account a non-zero probability of non-conforming packets. Copyright (C) 2002 John Wiley Sons, Ltd.