[Make-wifi-fast] Real router latency comparison

Dave Taht dave.taht at gmail.com
Sun May 17 14:57:37 EDT 2020


Tim:

well, can you run the test with the _be test, not mixed dscps, also?

Also are you getting actual latency differences from using each dscp?
or summing it?

A lot of vendors don't differentiate at all. individual stream plots
out of this data would show this, no difference in latency, across
mixed dscps. in that case, one vendor actually following the 802.11e
spec, and others not, hmmm.

But: I've always recommended openwrt turn off 802.11e differentiation
on the AP, and this is why. It is VASTLY better to optimumize for
802.11n aggregation than 802.11e. on the AP. Clients, sure, try it....

I know/knew exactly what the two things wrong here, but never got
around to fixing it, as in practice dscps are hardly used at all.

1) we use up a txop in all the queues rather than just 1 at a time.
2) the codel algorithm has some weird code in it the mucks with the
target when it shouldn't and we use a too large target for 5ghz.

Still not bad for a 2x ap against bigger fatter tech that I personally
have never seen.

THX FOR THE TEST!!!


On Sun, May 17, 2020 at 11:47 AM Tim Higgins <tim at smallnetbuilder.com> wrote:
>
> Hi all,
>
> Attached are previews of results I'll be publishing in the Part 2 article.
>
> The same test was run on four routers:
> - NETGEAR R7800 (Qualcomm 4 stream AC)
> - Evenroute IQrouter v3 (Mediatek 2 stream AC, Wifi stack is "vanilla"OpenWRT for the MT76" according to Evenroute)
> - NETGEAR RAX120 (Qualcomm four stream AX) - OFDMA enabled
> - NETGEAR RAX45 (Broadcom four stream AX) - OFDMA enabled
>
> AP set to Channel 36, 80 MHz bandwidth, WPA2 PSK connection
>
> Each STA shows 867 Mbps link rate. Signal level ~ -45 dBm.
>
> iperf3 TCP/IP traffic run simultaneously to all four STAs (Intel AX200, Win 10, 21.80.2.1 driver)
> bitrate  (-b): 50Mbps
> length (-l): 256 Bytes
> DSCP values (--dscp), one per STA: 0 (CS0), 96 (CS3), 160 (CS5), 192 (CS6)
>
> 200ms interval ping run concurrently from AP to STA on each pair. Ping is always AP to STA for both uplink and downlink traffic.
>
> Congestion was measured using a Qualcomm AX STA associated to the AP, running 1bps of traffic so that stats could be recorded.
>
> Thoughts?
>
> ===========
> Tim
> _______________________________________________
> Make-wifi-fast mailing list
> Make-wifi-fast at lists.bufferbloat.net
> https://lists.bufferbloat.net/listinfo/make-wifi-fast



--
"For a successful technology, reality must take precedence over public
relations, for Mother Nature cannot be fooled" - Richard Feynman

dave at taht.net <Dave Täht> CTO, TekLibre, LLC Tel: 1-831-435-0729


More information about the Make-wifi-fast mailing list