On the performance evaluation of an ATM self-routing multistage switch with bursty and uniform traffic

An N × N self-routing multistage switching network with bursty and uniform traffic, which is composed of S × S switching elements with output queuing and has K = log_s N stages, is considered. The input traffic is modeled by a two-state (active/silent) Markov chain. Each arriving cell selects one of the N outlets as its destination independently and with equal probability 1/N. The performance of the switching network is studied by means of a simulation method. An approximate traffic model of a four-state Markov chain is proposed to characterize the cell arrival processes to further stages. Comparison with the simulation results shows that the proposed traffic model gives a good approximation, which implies that the performance of the switching element mainly depends on the first two moments of the active and silent periods and is not sensitive to the shapes of their distributions. The influence of bursty traffic on the performance of the switching network is discussed and illustrated by some numerical examples. It is found that the performance of the switching network can be substantially improved by adding a simple traffic randomization mechanism