Performance evaluation of packet voice multiplexer with selective packet discarding control

This paper evaluates the performance of packet voice multiplexer with selective packet discarding (SPD) control. The control scheme is based on the embedded ADPCM coding technology wherein the encoded information is grouped into two separate parts: more significant part (MSP) and less significant part (LSP). When a certain congestion threshold is exceeded, the newly arriving LSP packets will be selectively discarded to relieve the congestion. By assuming that the packets arrive at the multiplexer in phase-type Markov renewal process (PH-MRP) and require the service times of phase-type (PH) with very large number of phases, we model the multiplexer under consideration as finite PH-MRP’[^x’]/PH/1/m(m1) queue with batch arrivals and control on the batch size. Such a queueing model is analyzed by using the matrix-analytic method, so that recursion formulae and their efficient computation algorithm for the queue length distribution, blocking probabilities and waiting time distribution are derived. Numerical examples demonstrate that the SPD control scheme indeed achieves significant performance improvement over uncontrolled system. [In Japanese.]