[Bloat] congestion control and video conferencing

[Dave Taht <dave.taht@gmail.com>]

re: http://conferences.sigcomm.org/sigcomm/2013/papers/fhmn/p21.pdf

"We have found that the algorithm works as expected when a GCC ow
accesses the bottleneck in isolation, whereas it is not able to
provide a fair band-
width utilization when a GCC flow shares the bottleneck with either a
GCC or a TCP flow."

I basically long ago came to the conclusion that video conferencing
can't work without aqm and packet scheduling on the bottleneck links.
Period. I wasn't aware there was also trouble even with two video
conferencing flows sharing the same link!

I run webrtc based and google hangout based talks through cerowrt with
fq_codel all the time while doing other benchmarks, but haven't come
up with good ways to quantify that. The results are generally pretty
spectacular with the few glitches I get attributable to wifi
misbehavior more than anything else.

Is anybody else testing webrtc through various aqm technologies?

I'd like to repeat this paper in a fq_codeled/pied/red'd environment
and see what happens... anyone interested in helping?


-- 
Dave Täht

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