Enhanced mechanisms for explicit congestion notification in TCP/IP networks

In this thesis the suitability of Backward Explicit Congestion Notification (BECN) for IP networks is evaluated extensively. The BECN mechanism has previously been used in non-IP networks, but there has been limited experimental investigation into the application of the BECN scheme as congestion control mechanism in TCP/IP networks. In this thesis, an enhanced algorithm for BECN which uses Internet Control Message Protocol (ICMP) Source Quenches (ISQs) for backward notification in IP networks is considered. A mathematical model is developed to analyze and predict the throughput and queue variation for TCP flows in Explicit Congestion Notification (ECN) and BECN capable networks while comparative performance evaluation of Random Early Detection (RED), ECN and our enhanced BECN mechanism is undertaken to validate the model. This thesis also undertakes performance evaluation of a suggested Multi-level Backward Explicit Congestion Notification (MECN) mechanism in comparison to the ECN and BECN mechanisms with long-lived TCP traffic.; The impact of two-way traffic and a lossy reverse path on performance of the ECN and BECN mechanisms is extensively investigated and results showed that two-way traffic and loss on the reverse path can result in significant queue fluctuations for BECN due to loss of ISQ messages. A lossy reverse path did not significantly affect queue fluctuations for ECN due to reliable delivery of congestion notification via ECN Echo ACKs (ECE-ACKs).; Deductions from these results led to the proposal and evaluation of a novel algorithm (ECN+BECN mechanism) which combines the merits of the ECN and BECN mechanisms. Experimental results show that the combined mechanism benefits from BECN's early notification under heavy congestion and ECN's reliable delivery of congestion notification. It is also shown that the ECN+BECN scheme can significantly reduce the ICMP Source Quench reverse traffic in a network compared to the BECN scheme.