A TRAFFIC MEASUREMENT-BASED MODELING OF SUPERPOSED ATM CELL STREAMS

Superposed ATM cell streams have burstiness and strong autocorrelation properties. This paper investigates traffic measurement-based modelin g method for superposed ATM cell streams. We develop a new measurement method based on monitoring both the waiting time distribution in a mo nitoring queue and the autocorrelation of cell interarrival times. Thr ough the monitoring queue, we directly observe the queueing effect of superposed cell flows on ATM multiplexers. The measured traffic is mod eled as the two-state MMPP. With the measured traffic, we estimate the cell loss probability in ATM multiplexers from the MMPP/D/1/K queue. Our method successfully works with homogeneous and heterogeneous super position of traffic sources including voice, data, and video. These re sults can be applied to the evaluation of ATM multiplexers, traffic en gineering, and network performance monitoring.