Search RFCs

RFC Editor

File formats:

Status:

EXPERIMENTAL

Authors:

C. Raiciu
M. Handley
D. Wischik

Stream:

IETF

Source:

mptcp (tsv)

Cite this RFC:TXT  | XML  |  BibTeX

DOI:  https://doi.org/10.17487/RFC6356

Discuss this RFC: Send questions or comments to the mailing listmultipathtcp@ietf.org

Other actions:View Errata  | Submit Errata  | Find IPR Disclosures from the IETF  | View History of RFC 6356

Abstract

Often endpoints are connected by multiple paths, but communicationsare usually restricted to a single path per connection. Resourceusage within the network would be more efficient were it possible forthese multiple paths to be used concurrently. Multipath TCP is aproposal to achieve multipath transport in TCP.

New congestion control algorithms are needed for multipath transportprotocols such as Multipath TCP, as single path algorithms have aseries of issues in the multipath context. One of the prominentproblems is that running existing algorithms such as standard TCPindependently on each path would give the multipath flow more thanits fair share at a bottleneck link traversed by more than one of itssubflows. Further, it is desirable that a source with multiple pathsavailable will transfer more traffic using the least congested of thepaths, achieving a property called "resource pooling" where a bundleof links effectively behaves like one shared link with biggercapacity. This would increase the overall efficiency of the networkand also its robustness to failure.

This document presents a congestion control algorithm that couplesthe congestion control algorithms running on different subflows bylinking their increase functions, and dynamically controls theoverall aggressiveness of the multipath flow. The result is apractical algorithm that is fair to TCP at bottlenecks while movingtraffic away from congested links. This document defines an Experimental Protocol for the Internet community.

For the definition ofStatus,seeRFC 2026.

For the definition ofStream, seeRFC 8729.