CeroWrt Triumphs over Bufferbloat

Posted on Mar 10, 2015 by Rich Brown.

The CeroWrt project, and its implementation of fq_codel as seen in the current build of the firmware, eliminates the problem of bufferbloat. These changes have been pushed into Linux kernel and the OpenWrt mainline (“Barrier Breaker” release), and are now widely available.

Bufferbloat is the undesirable latency that comes from a router or other network equipment buffering too much data. It has plagued network routers from the early days. The problem was made worse as RAM became cheaper: network engineers worried that dropping packets would make the network slow, so there was an incentive to buffer more and more packets. This had the paradoxical effect of retaining too many packets, which hold up all the traffic behind those buffers.

Many efforts through the 1990s and 2000’s attempted to address the problem. Random Early Drop (RED) and its variants showed promise, but didn’t monitor the proper variables, and were thus hard to configure properly and would hurt performance if not tuned correctly. Various quality of service (QoS) policies can give priority to certain types of traffic, but they’re hard to configure. As traffic types change and evolve, these policies become a maintenance hassle, since they need to be rewritten on a regular basis.

In early 2012, Kathie Nichols took another look at the problem of overbuffered routers and designed the CoDel (pronounced “coddle’) algorithm. The major insight was that the best way to avoid “too much buffering” was to monitor a packet’s sojourn time - the time elapsed between when it was queued for transmssion and dequeued. If that time exceeds a certain threshold (generally 5 msec), it indicated that the packet had been queued for a long time. CoDel would then drop a percentage of those packet to provide feedback to the sender that it was using more than its share of the available capacity. An elaboration to the CoDel algorithm - fq_codel from Eric Dumazet - placed packets for each source/destination flow in a separate queue, and applied the CoDel algorithm to each queue to extremely good effect.

The resulting fq_codel qdisc was put in to the Linux 3.5 kernel in July 2012.

RSS feed

Recent News & Articles

Sep 6, 2018 Wiki page
Pete Heist's Thoughts on ECN
Sep 5, 2018 Wiki page
Dave Taht's Stance on ECN
Sep 4, 2018 Wiki page
Jonathan Morton's Take on ECN
Sep 3, 2018 Wiki page
ECN-Sane Project
Aug 24, 2018 Wiki page
ECN-Sane Project

Find us elsewhere

Bufferbloat Mailing Lists
#bufferbloat on Twitter
Google+ group
Archived Bufferbloat pages from the Wayback Machine

Sponsors

Comcast Research Innovation Fund
Nlnet Foundation
Shuttleworth Foundation
GoFundMe

Bufferbloat Related Projects

Congestion Control Blog
Lede Project (OpenWrt)
Flent Network Test Suite
Sqm-Scripts
The Cake shaper
AQMs in BSD
IETF AQM WG

Network Performance Related Resources


Jim Gettys' Blog - The chairman of the Fjord
Toke's Blog - Karlstad University's work on bloat
Voip Users Conference - Weekly Videoconference mostly about voip
Candelatech - A wifi testing company that "gets it".