Dynamic Flow Control in Store-and-Forward Computer Networks

In a recent paper we presented an analysis of flow control in store-and-forward computer communication networks using a token mechanism. The analysis assumed equilibrium conditions for a selected set of system parameters which were not dynamically adjusted to stochastic fluctuations in the system load; this mechanism was referred to as "static flow control." In this paper we study a "dynamic flow control" in which parameters of the system are dynamically adjusted to match the availability of resources in the network. Based on Markov decision theory, an optimal policy to dynamically select the number of tokens is formulated. Because an exact solution to the problem is extremely difficult, an effective heuristic solution to the problem is presented. Numerical results are given and it is shown that the throughput-delay performance of a network is better with dynamic control than with static control.