Extending a Control System Model for Rate-based Congestion

The Asynchronous Transfer Mode (ATM) is a network standard currently under analysis and development. The ATM Forum, a consortium undertaking the majority of the devel-opment of ATM protocols, proposed guidelines for congestion control which include a rate-based control mechanism. Some of the early work in this field made use of linear con-trol theory to develop rate-based congestion control algorithms. The models established in this early work are inadequate for networks comprised of multiple switch elements. This thesis extends the model to more complicated networks. The effect of adding link delays to a one switch system is first analyzed. Then, the model for a network of switches is created. This model requires the representation of several dynamics not present in a one switch system. These dynamics result from the influences that networked switches have on one another. For example, the queue of any switch can impact the flow of cells through-out the network because queues cause delays and restrict the flow of cell traffic. Another previously unmodeled dynamic results from the way in which the rate-based congestio