Cell loss probability of a finite ATM buffer queue
Various formulations have been proposed to analyse the statistical behaviour of multimedia traffic over an ATM switch of infinite capacity. The authors present different approaches for calculating the cell loss probability of a finite buffer queue. They first model an ATM traffic source using a two-state Markov-modulated Bernoulli process (MMBP), and then calculate the cell loss probability of a finite ATM buffer queue fed with N such traffic sources using different approaches. Numerical comparison confirms that all these approaches yield similar results. The first two approaches are based on the matrix geometric formulation. The authors derive the queue length distribution of a finite ATM buffer and solve for the state probability. The cell loss probability is then approximated by the saturation probability (i.e. the probability that the buffer is full, in the first approach). In the second approach, having obtained the state probability, the authors derive the exact expression for the cell loss probability using conditional probability. The third approach is based on an approximation for the buffer content of a queueing buffer. The cell loss probability is formulated using a generating function and solved by examining the smallest pole of its function.
