YACAD: Yet Another Congestion Avoidance Design for ATM-based networks

This dissertation proposes YACAD (Yet Another Congestion Avoidance Design for ATM-based Networks), a congestion prevention model that includes admission control, traffic shaping, and link-by-link flow control for ATM-based networks. Network traffic in this model is composed of real-time traffic and data traffic. As real-time traffic is delay-sensitive and connection-oriented, its call acceptance is based upon the effective bandwidth at all nodes. Effective bandwidth is defined as a vector of bandwidth and maximum node delay. As data traffic can be either connection-oriented or connectionless, it is subject to link-by-link flow control based on a criterion known as effective buffer which is defined as a scalar of buffer size. Data traffic is not delay-sensitive but is loss-sensitive. Traffic shaping is imposed on real-time traffic to ensure a smooth inflow of real-time cells. YACAD also allocates a large buffer (fat bucket) to data traffic to accommodate sudden long bursts of data cells. Absence of data cell loss is a major feature of YACAD. Two simulation studies on the performance of the model are conducted. Analyses of the simulation results show that the proposed congestion avoidance model can achieve congestion-free networking and bounded network delays for real-time traffic at high levels of channel utilization. The maximum buffer requirements for loss-free cell delivery for data traffic, and the cell loss probabilities for real-time traffic are also obtained. In addition, results of performance comparisons to other similar models have shown that YACAD outperforms several other leaky-bucket based congestion control methods in terms of cell loss probability for real-time traffic. The simulation source program has also been verified using existing queueing theories, and the Paired-t Confidence Interval method with satisfactory results at 99% confidence level.