In this thesis, the problem of congestion control in computer communication networks is studied. The method employed consists of a combination of a dynamic routing policy that minimizes the delay, and buffer management schemes to control the flow of data. These schemes include (1) Complete Sharing, (2) Square Root Sharing, and (3) Sharing with Minimum Allocation.

Document studies the broad area of flow control and sequencing in computer communication networks. A special use of message sequencing is shown to reduce to a Markov process. A more general sequencing model is also studies; only approximate results are obtained in this case. The author formally shows that an end-to-end flow control scheme in computer communication networks can be modeled as a closed network of queues. He also introduces and analyzes a flow control scheme based on limiting the permit generation rates. It turns out that this scheme is very effective. More importantly, this scheme involves essentially no computational complexity. This allows us to measure and adjust the network performance with little difficulty. The simplicity also allows us to incorporate more realistic aspects of network protocols such as piggy-back acknowledgement, time-out and gate functions into our analysis. Two simple but effective distributed dynamic flow control schemes are proposed and analyzed. It is shown that with either of these two schemes, a network can respond well to the dynamically changing environment.

(no abstract)