Experimental Specification of New Congestion Control Algorithms

Date: Jul 2007

This document describes the process that the IETF transport area directors employ when new congestion control algorithms that differ significantly from currently-published IETF standards are brought to the IETF for specification and publication as Experimental RFCs.

1. Introduction

For a number of years, there has been considerable interest in the research community on new congestion control algorithms that diverge from the congestion control standards specified in the IETF, for which [RFC2914] outlines the principles.

One current motivation for such new congestion control algorithms [I-D.ietf-tsvwg-cc-alt] is better support for high-speed paths. Research in this area has resulted in a number of promising congestion control variants. Bandwidths that have been considered "high-speed" years ago are now readily available to consumers in many countries, resulting in a corresponding interest by the operating system community to deliver mechanisms to the users that can efficiently use these kinds of network paths. Several widely- deployed operating systems already implement non-RFC-compliant congestion control mechanisms for this purpose. This is happening without much community review and often without a publicly available technical specification, which is troubling.

The IETF transport area offers to be a venue for community discussion and review, followed by an eventual technical specification, of new congestion control variants that go beyond the IETF's general congestion control principles [RFC2914].

Section 2 describes the process that the IETF transport area directors employ when new congestion control algorithms are brought to the Transport Area for specification and publication as Experimental RFCs. The main feature of this process is an expert review phase performed by the IRTF's Internet Congestion Control Research Group (ICCRG). The decision of whether a specific new congestion control schemes goes beyond the principles of [RFC2914] and hence will undergo the process described in this document lies with the transport area directors, who will use the expertise of the transport area directorate and other congestion control experts to aid their decision.

The wider transport community established consensus on this process after the presentation and discussion during the ICCRG meeting in February 2007 and the TSV area meeting during IETF-68 in March 2007. At the time of writing, there are at least three congestion control algorithms that have been proposed for Experimental specification through this process.

2. Experimental Specification of New Congestion Control Algorithms

The main goal of this process is to publish a number of thoroughly- analyzed congestion control variants as Experimental RFCs. These documents will be published with statements that indicate that the IETF believes them to be safe for experimentation on the Internet, or safe for experimentation in certain more restricted network environments. This is similar, for example, to the statements the IETF made for QuickStart [RFC4782] or HighSpeed TCP [RFC3649].

To be able to make such statements about Experimental congestion control variants, a review phase needs to occur in the community. The transport area directors will use the expertise in the IRTF's Internet Congestion Control Research Group (ICCRG) to perform an expert review during the initial phase of this process. The proposers of high-speed congestion control variants will present their mechanism and results to the ICCRG, which will analyze and discuss these variants with the goal of determining whether they can be declared safe for experimentation, and under which conditions. [I-D.irtf-tmrg-metrics] defines some metrics that will be useful during this expert review. [I-D.ietf-tsvwg-cc-alt] describes the consensus of the IETF transport area on the best current practice in evaluating new congestion control algorithms, which the ICCRG review should consider.

The ICCRG will document the results of their expert review in a write-up, which will accompany a new congestion control algorithm when it is proposed for follow-on technical specification in the IETF's transport area. Current consensus is to add a work item to the charter of the TCP Maintenance and Minor Extensions (TCPM) working group for the Experimental standardization of such ICCRG- reviewed congestion control variants, because all current proposals in this space are TCP modifications. In the future, new congestion control variants for other protocols may become work items of other working groups.

When new congestion control algorithms that significantly differ from the general congestion control principles outlined in [RFC2914] are brought to the IETF transport area without a prior expert review by the ICCRG, the transport area directors will request the ICCRG to perform such a review as an initial step. The RFC Editor is encouraged to do the same for independent submissions [I-D.iab-rfc-independent].

3. Acknowledgments

Thanks to Mark Allman, Brian Carpenter, Ted Faber, Sally Floyd, Cullen Jennings and Michael Welzl for their comments on this document.

4. References

[I-D.iab-rfc-independent]

Klensin, J. and D. Thaler, "Independent Submissions to the RFC Editor", draft-iab-rfc-independent-00 (work in progress), March 2007.

[I-D.ietf-tsvwg-cc-alt]

Floyd, S. and M. Allman, "Specifying New Congestion Control Algorithms", draft-ietf-tsvwg-cc-alt-04 (work in progress), June 2007.

[I-D.irtf-tmrg-metrics]

Floyd, S., "Metrics for the Evaluation of Congestion Control Mechanisms", draft-irtf-tmrg-metrics-09 (work in progress), March 2007.

[RFC2914]

Floyd, S., "Congestion Control Principles", BCP 41, RFC 2914, September 2000.

[RFC3649]

Floyd, S., "HighSpeed TCP for Large Congestion Windows", RFC 3649, December 2003.

[RFC4782]

Floyd, S., Allman, M., Jain, A., and P. Sarolahti, "Quick- Start for TCP and IP", RFC 4782, January 2007.

Appendix A. Informational Documentation of Implementation Choices

The IETF encourages vendors of network stacks to describe their implementation choices for the information of the community, and offers to publish such reports as Informational RFCs. Such documents will come with a clear statement calling out that the contents of these Informational drafts document the implementation choices of particular stacks and are not recommendations, endorsements, etc. by the IETF.

Earlier, this class of documents was discussed as part of the process in Section 2. Further reflection determined that such documents do not in fact require a new process and can be brought directly to the relevant working groups of the IETF. This appendix was added to clarify this reconsideration.

Author's Address

Lars Eggert
Nokia Research Center
P.O. Box 407
Nokia Group 00045
Finland

Phone: +358 50 48 24461
Email: lars.eggert@nokia.com
URI: http://research.nokia.com/people/lars_eggert/