[Make-wifi-fast] Resolving Bufferbloat in TCP Communication over IEEE 802.11n WLAN by Reducing MAC Retransmission Limit at Low Data Rate
dave.taht at gmail.com
Fri Sep 8 13:41:22 EDT 2017
It is nice to see these lists are not entirely dead. I've spent the
last few weeks catching up on published bufferbloat research on google
scholar and there has been more than a few papers worth reading, but
only this one on wifi (besides "ending the anomaly") stood out.
I liked this paper because we'd discussed weakening mac retransmit
functions in the past, and they went ahead and did it, with an
admittedly simple bimodal function between 80mbits and below. I was
lost in the weeds (whenever this was last discussed) trying to come up
with an integrated scheme that leveraged apparent load against
retransmits, and pulled in minstrel itself in conjunction with
fq_codel to try and have an appropriate retransmit figure - the core
minstrel insight was that you could, indeed, retry a lot at high
rates, but the bloat insight was that you really didn't want to do
that at lower ones.
I'm still lost in the weeds.
More information about the Make-wifi-fast