Statistical multiplexing with arbitrary on/off sources

In an ATM environment many types of traffic, such as voice, data, and video, are to be efficiently transported by the same network. An ATM multiplexer receives cells (fixed size packets of 53 octets length) from a number of different incoming links and then transmits them out onto a single outgoing link. A finite buffer is provided in the multiplexer to accommodate the multiple arrivals of cells. We develop a methodology for characterizing the superposition process of N/spl ges/2 discrete-time arbitrary on/off sources. The superposition is a discrete-time semi-Markov process with O(2/sup 2N/) states. We use the superposition model to analyze a finite-buffer statistical multiplexer with multiple arbitrary on/off input sources. We study the effect of various traffic parameters on the queueing performance.