The Entropy of Tra c Streams in ATM Virtual

We model an ATM virtual circuit as a tandem queueing system. At each queue, transmission of the virtual circuit traac is interrupted by cross traf-c, causing the scattering and coalescing of cells. We study this phenomenon using entropy as a traf-c descriptor. We compute the entropy of traac streams produced in simulation, via an estimation technique based on the Lempel-Ziv universal data compression method. The estimator modiies the Lempel-Ziv method to compute entropy rather than compress data. We show that the entropy of virtual circuit traac at successive queues can either increase or decrease depending upon the types of input traac. Even when bursty input is scattered, its entropy does not achieve maximum entropy within a reasonable number of queues. We also deene a distance metric to compare the correlation structures of two output processes and observe this metric at successive queue outputs.