Performance analysis of large-scaled voice/data multiplexers

Abstract

The main objectives of this dissertation work are to analyze the throughputs of a packet-switched network with error/flow control schemes over a noisy channel, and to investigate the queueing behaviors of a large-scaled integrated voice/data multiplexer. For the case that transmission errors occur in bursts with the transmission conditions of data blocks correlated, we investigate the throughput efficiency of the system that utilizes the go-back-N(GBN) and stop and wait (SW) automatic repeat request (ARQ) strategies, based on the renewal theorem for a regenerative process when forward and backward channels both have Markov error characteristics with dependency between messages. As a result, simple analytical formulas for the throughput efficiency are obtained. Particularly, in the case of the GBN ARQ scheme, we show that the result of throughput analysis can be extended to a general form that includes Leung``s result. For high error-rate channels, we show that a Markovian system with highly correlated errors yields better throughput efficiency than the system with random errors. In addition, we show that the case in which the error pattern clustered in one channel makes the system throughput inferior to the case that errors are distributed in forward and backward channels. Next, we are concerned with the performance analysis of a packet network with sliding-window flow control and selective repeat ARQ schemes. In this delayed feedback queueing analysis, exact analysis is mathematically very complex except for a few cases in which inter-arrival and service time are exponentially distributed. The difficulty of performance analysis lies mainly in the fact that the input processes of both queues are neither Poisson nor renewal process even if the arrival process is Poisson. To get a closed form solution, we approximate the original system by a tandem queueing system with the characteristics of an imbedded Markov renewal process in the steady state. Also, we investi...