dtaht

fixes for bufferbloat  pending in DOCSIS 3.1

Discussion created by dtaht on Sep 1, 2014
Latest reply on Sep 2, 2014 by [shaw]ali
Branched from an earlier discussion

I'd like to point out that while the cable industry has fixes for bufferbloat on cable modems pending in DOCSIS 3.1, that the aftermarket home router firmware makers have all adopted fq_codel to make things MUCH better if you put it in front of an existing cable modem to better manage the bandwidth, both up and down. So far as I know the cable industry has NOT come up with a way to better manage the downstream bandwidth as yet, although arris has published some interesting info.... (  http://snapon.lab.bufferbloat.net/~d/trimfat/Cloonan_Paper.pdf ) based on their ns2 models. (in contrast, nearly all the other benchmarks I've referenced here are against real world implementations on real world routers against real world cable services)

 

openwrt barrier breaker, cerowrt, dd-wrt, ipfire, pfsense, etc, all now have code that can better manage both up/down bandwidth and latency, if you are willing to expend 50-100 bucks on a router that has the code. Also products with "Streamboost" are fq_codel based. More are appearing daily.

 

Some benchmarks against cable devices, using these new QoS/Fair Queueing/AQM systems, including a benchmark of pie:

 

Burnt Chrome: May 2014

 

Getting SQM running right on Jim Reisert's network with CeroWrt 3.10

 

The relevant ietf internet drafts on standardizing these technologies are here:

 

draft-white-aqm-docsis-pie-01

draft-hoeiland-joergensen-aqm-fq-codel-00 - FlowQueue-Codel

https://datatracker.ietf.org/doc/draft-nichols-tsvwg-codel/

Outcomes