[Make-wifi-fast] tcp_pacing_rate_shift and smaller txops

[bkil]

Interesting observation. Do you think the new wmm rules could prevent
fine tuning of these constants later on?

https://git.kernel.org/pub/scm/linux/kernel/git/sforshee/wireless-regdb.git/commit/?id=fde57be08890095132b9f7d28aaaf9a026b2e608

On Mon, Jul 2, 2018 at 9:44 PM, Dave Taht <dave.taht at gmail.com> wrote:
> wifi APs have the ability to advertise in the beacon, the actual
> maximum time of a txop for clients, and 802.11n, for example, cannot
> queue an aggregate in the VO queue. the otherwise lovely
> tcp_pacing_rate shift thing sets this to a fixed 4ms, when perhaps it
> could leverage the actual setting for the queue in question, at least,
> as defined by the beacon or 802.11e standard.
>
> I've long thought that perhaps a gaming wifi router would advertise
> smaller txops, and for all I know, some do.
>
> this is currently controlled by cwmin and cwmax parameters in hostapd
>
> # Best effort
> #tx_queue_data2_aifs=3
> #tx_queue_data2_cwmin=15
> #tx_queue_data2_cwmax=63
> #tx_queue_data2_burst=0
>
> I did a lot of fiddling with this (lowering it to 1,1.5,2ms,3ms,4ms)
> at one point (that I cannot currently find). It hurt bandwidth but
> helped latency, a lot, and this was from before we shipped fq_codel
> for wifi.
>
> --
>
> Dave Täht
> CEO, TekLibre, LLC
> http://www.teklibre.com
> Tel: 1-669-226-2619
> _______________________________________________
> Make-wifi-fast mailing list
> Make-wifi-fast at lists.bufferbloat.net
> https://lists.bufferbloat.net/listinfo/make-wifi-fast