Skip to main content

11.1 Related Work Evaluating ECN

This section discusses some of the related work evaluating the use of ECN. The ECN web page [ECN] contains pointers to additional papers as well as ECN implementations.

[Floyd94] considers the advantages and disadvantages of adding ECN to the TCP/IP architecture. As shown by simulation-based comparisons, one advantage of ECN is avoiding unnecessary packet drops for short connections or delay-sensitive TCP connections. A second advantage of ECN is avoiding some unnecessary retransmission timeouts in TCP. This paper discusses in detail the integration of ECN with TCP congestion control mechanisms. Possible disadvantages of ECN discussed in this paper are that non-compliant TCP connections might falsely advertise themselves as supporting ECN, and that TCP ACK packets carrying the ECN-Echo message might themselves be dropped in the network. The first of these two issues is discussed in the appendices of this document, and the second is addressed by adding the CWR flag in the TCP header.

Experimental evaluations of ECN include [RFC2884, K98]. The conclusions of [K98] and [RFC2884] are that ECN TCP gets modestly better throughput than non-ECN TCP; ECN TCP flows are fair to non-ECN TCP flows; and ECN TCP is robust to bidirectional traffic (bidirectional congestion) and to multiple congested gateways. Experiments with many short web transfers show that while most short connections have similar transfer times with or without ECN, a small fraction of short connections have very long transfer times for the non-ECN experiments compared to the ECN experiments.