A two-node asynchronous transfer mode (ATM) network is considered. Several video sources are assumed multiplexed at the source node. Five alternative algorithms for bandwidth allocation to video sources are considered using: static-slot assignment, buffer-population-based dynamic slot assignment, first-come-first-serve slot assignment and rate-based dynamic slot assignment. The average cell delay and the coefficient of variation of the delay for the five algorithms are determined by computer simulation and compared. It is concluded that the adaptive slot assignment algorithm and the buffer-population-based dynamic slot assignment algorithm are suitable for packet video in ATM networks.>

In some computer network architectures a process communicates with another process across the network by establishing a virtual circuit. Consider a virtual circuit from node S to node D connected by multiple links. The delay suffered by the packets of the virtual circuit has three components: waiting time or queueing delay, service time or transmission delay, and resequencing delay. The sum of these is called the total delay. The author presents the distribution of the total delay for three different queuing models of node S: G/M/m model, G/M/ infinity model, and M/H/sub K// infinity model.>

The relationship between the mean resequencing delay and variations in packet transmission times is studied. Assuming a Poisson stream of packets, a K-stage hyperexponential transmission time distribution and an infinite number of equal capacity links connecting the source and destination nodes, the authors derive an expression for the mean resequencing delay. This result provides an upper bound on the mean resequencing delay for nodes connected by finitely many links. They observe that for the two-stage and three-stage hyperexponential distribution, the mean resequencing delay varies almost perfectly linearly with the squared coefficient of variation. An asymptotic bound analysis can explain this behavior.>