A discrete-time performance model for integrated service ATM multiplexers

Voice sources are modeled as discrete-time on-off processes; video sources with uniform activity level are modeled by a discrete-time birth-death process; and, the arrival process from an aggregation of data sources is assumed to form a Bernoulli process with a general bulk size distribution. The integration of these processes belongs to a family of processes, the discrete-time batch Markovian arrival processes (D-BMAP). An ATM multiplexer whose input consists of a mix of the different types of traffic described above can therefore be modeled as a D-BMAP/D/1 queueing system. The authors investigate an approximation which matches the original traffic model to a two-state D-BMAP. This procedure results in greatly reduced complexity allowing it to be used for network traffic control while retaining accurate performance prediction