Rate-Based Flow Control

On high-speed networks, the round trip delay from source to intermediate node and back is relatively large compared with packet transmission times; thus, the window-based method of flow control does not react quickly enough to prevent congestion and large delay. An alternative method is to employ a rate-based flow control scheme, in which each node is assigned a maximum rate and an average rate of transmission. The purpose of this thesis is to develop a criterion for evaluating how well a given assignment of rates will perform in preventing congestion in the network. The criterion to be used is the probability that the instantaneous aggregate arrival rate at any one node is greater than the capacity of the outgoing link of that node. In order to simplify the development of an expression for this probability, an on/off model is used for the sources, and only a subset of the network is considered a set of sources feeding into one intermediate node with an outgoing link. This thesis examines how the assigned rates and the burstiness of the sources affect this probability, and analyzes the effect of adding a new source to the system.