An ATM statistical multiplexer with on/off sources and spacing: numerical and analytical performance studies

This paper analyzes the performance of an ATM statistical multiplexer fed by randomly spaced on / off sources, for both finite and infinite buffer-size cases. In the underlying randomly spaced on/off traffic model, the number of cells generated in an ON period and the length of an OFF period are both geometrically distributed. The interarrival time between two adjacent cells generated in an ON period is also assumed to be geometrically distributed (i.e., the cells are randomly spaced). In the paper, a numerical algorithm is presented to obtain the cell loss probability for the finite buffer-size case. By means of a generating-functions approach, an analytical method is developed to obtain the mean and the tail distribution of the buffer contents for the infinite buffer-size case. pact of the traffic burstiness and traffic smoothing (or spacing) on the queueing performance of the multiplexer is illustrated by some numerical examples. Furthermore, the study shown that the randomly spaced on/off traffic model can give a good upper-bound approximation for the fixedly spaced on/off (or sporadic) source model. Also presented in this paper is a new parameter matching method based on the first 2 moments of the interarrival time distribution and the decay rate of the tail distribution of the buffer contents, which results in a better approximation.