Modeling of aggregated bursty traffic sources in ATM multiplexers

A new model is developed to characterize the behavior of aggregated input traffic at a finite-buffer statistical multiplexer and to obtain the cell loss probability and queuing delay. The sources at the inputs of the statistical multiplexer alternate between active (on) and idle (off) periods, and are represented by their peak and average transmission rates and the distribution of the lengths of their active and idle periods. The aggregate arrival process from these sources is modeled by a four-state Markov modulated Poisson process (MMPP). The parameters of the MMPP model are derived by considering the underload (UL) and overload (OL) period statistics of the aggregate arrival process. The mean lengths of the states, the average arrival rates in these states, and the transition probabilities among states are obtained by matching the mean and variance of the UL and OL period lengths; the mean and variance of excess arrivals in OL; and the mean and variance of excess service in UL of the MMPP and the actual arrival process.<
>