Performance analysis of an ATM multiplexer with different classes of bursty sources

An ATM multiplexer with different classes of bursty sources is considered for evaluation of the effects of bursty traffic variations on its performance. A discrete-time Markov chain model is developed to obtain cell loss probability both for the whole set of sources and for each class. Calculations carried out for the particular case of two classes, with a first class behaving as declared at call set up and a second class violating that reference situation by variation of some of the bursty traffic parameters. The results show that the most dangerous violation is represented by burst length increase, which will play a key role for the definition of congestion avoidance strategies