Modeling and dynamic scheduling of a queueing system with blocking and starvation

Consideration is given to the problem of dynamically controlling a computer communication network consisting of N stations that compete for the use of a single channel. The channel is needed by the stations in order to transfer the packetized information they got from independent sources to a central storage device. The stations, which have some local storage capacity, are modeled as finite queues fed by independent Poisson streams and the channel as a single exponential server. The performance objective is to avoid situations in which any of the queues is fully (blocking). The authors also consider systems in which the queues are fed by the server and the objective is to avoid situations in which any of the queues is empty (starvation), and a hybrid system, consisting of both types of queue. Many computer and communication systems fall within the framework of these models. The goal is to get a good control policy for these systems. The authors first prove certain structural properties of the optimal solution for the case N=2. These properties lead them to conclude that it is unlikely that a closed-form expression for the optimal policy could be found, but at the same time guide the derivation of a heuristic decision rule