[Cerowrt-devel] ietf report (Brief)

[Cerowrt-devel] ietf report (Brief)

[Dave Taht]

I think I should take the time to write down and sort out all that
happened at ietf. But that will take time I don't have (I dictated)
right now.

The big thing worth watching is bruce schneier's plenary talk...

http://recordings.conf.meetecho.com/Recordings/watch.jsp?recording=IETF88_TECH_PLENARY&chapter=part_7

I am increasingly annoyed about the actions of various intelligence agencies.

Some highlights from the aqm wg.

1) The pie folk announced the selection of a variant of pie as the
default and mandatory *on* aqm algorithm for docsis 3.1 cable modems.
There is support for alternate algos, and nothing has been decided for
CMTSes or docsis 3.0. It's going to take years for hardware to appear
on this...

I am very happy for something to be defaulting on.

2) Jim and I gave a preso based on cero + netem in the yurtlab
documenting some of the work we've been doing to evaluate fq_codel vs
pie vs drop tail

http://www.ietf.org/proceedings/88/slides/slides-88-aqm-3.pdf

(I really, really, really appreciate you all helping to make cero
stable enough to generate data we can trust on this and the atm fixes
in particular)

At RMCAT I got volunteered to look hard at the various proposals for
webrtc congestion control. If you haven't fiddled with it yet, load up
a current chrome or firefox web browser and tr
At ICCRG the presentation that grabbed my eye was google's on paced
tcp and the tcp reordering fixes. In particular I hope the reordering
stuff at the tcp layer gets rid of the need for it to happen at the
wireless later. I look forward, very much, to ripping that code out.

At TSVWG the Quic presentation was awesome. I got a chance for a more
private and in-depth brief the night before from jim roberts. He has
much passion and much data.... I encourage y'all to turn it on in
chrome and try it....

You can browse through the proceedings here:

https://datatracker.ietf.org/meeting/88/materials.html


-- 
Dave Täht

Fixing bufferbloat with cerowrt: http://www.teklibre.com/cerowrt/subscribe.html

Re: [Cerowrt-devel] ietf report (Brief)

[Török Edwin]

On 11/14/2013 10:21 PM, Dave Taht wrote:
> http://www.ietf.org/proceedings/88/slides/slides-88-aqm-3.pdf

Nice presentation!
For some reason I missed the announcement of the sysctl mentioned on page 10 (net.core.default_qdisc),
as it wasn't on http://kernelnewbies.org/LinuxChanges :(

Best regards,
--Edwin

Re: [Cerowrt-devel] ietf report (Brief)

[Maciej Soltysiak]

> 1) The pie folk announced the selection of a variant of pie as the
> default and mandatory *on* aqm algorithm for docsis 3.1 cable modems.
> There is support for alternate algos, and nothing has been decided for
> CMTSes or docsis 3.0. It's going to take years for hardware to appear
> on this...
Plus, there are other issues. For example, the docsis 3.0 cable modem
I got from my ISP is a Cisco EPC 3925 that doesn't seem to be very
hackable and there's just the Web GUI that gets the user/pass
overwritten after some time after resetting it. Not much to tinker
though.

Question: I'm running cero with the 90% qos throughput trick behind
this Docsis 3.0 modem. Does it look like in the future I should
replace it with a 3.1 modem with PIE and stop doing the 90% trick on
cero?

Best regards,
Maciej