Dynamic control of session input rates in communication networks

We consider a distributed iterative algorithm for dynamically adjusting the input rate of each session of a voice or data network using virtual circuits so as to exercise flow control. Each session origin periodically receives information regarding the level of congestion along the session path and iteratively corrects its input rate. In this paper we place emphasis on voice networks, but the ideas involved are also relevant for dynamic flow control in data networks. The algorithm provides for the addition of new and termination of old sessions and maintains at all times feasibility of link flows with respect to capacity constraints. Fairness with respect to all sessions is built into the algorithm and a mechanism is provided to control link utilization and average delay per packet at any desired level.