HIERARCHICAL SCHEMES FOR ROUTING AND FLOW CONTROL IN COMMUNICATION NETWORKS.

This dissertation describes three different hierarchical schemes for routing and flow control in communication networks. The basic idea in these schemes is to distribute the decision making capability among different hierarchical levels. A part of the decision is made at the lower level of network nodes and another part is made at the higher level of Coordinators. Specific problems at the lower level of nodes and the higher level of Coordinators are formulated in an optimization framework and solutions to these problems result in the overall control scheme. Two modes of operation for the Coordinators are identified and a step by step procedure to implement these modes of operation is also developed. Performance evaluations of these hierarchical schemes are conducted using stochastic models and simulation experiments. Different models for the nodes are used while developing these hierarchical schemes. In one of these schemes a queueing model is used to design a buffer allocation scheme to implement flow control while in another a state dependent model is used to design a combined routing and flow control scheme. The solution to this combined problem is obtained through the concepts of system stabilization. Major strong points of these hierarchical schemes are as follows. They provide a framework for integrating routing and flow control functions. They provide consideration of multiple objectives such as delay, throughput, and fairness individually. By implementing the Coordinators action at a slower time scale, the trade-off existing in information overheads and optimality can be exploited