[Make-wifi-fast] Weird periodic latency

[Michael Yartys]

On Friday, January 15th, 2021 at 20:10, Dave Taht <dave.taht at gmail.com> wrote:

> oops meant to cc the list
>
> On Fri, Jan 15, 2021 at 11:09 AM Dave Taht dave.taht at gmail.com wrote:
>
> > On Fri, Jan 15, 2021 at 9:00 AM Michael Yartys via Make-wifi-fast
> >
> > make-wifi-fast at lists.bufferbloat.net wrote:
> >
> > > I assume that I got it right by running:
> > >
> > > flent --test-parameter ping_hosts=1.1.1.1 -l 300 netperf-eu.bufferbloat.net tcp_8down
> > >
> > > Either way, I think I'll have to do this test at another time since someone is streaming TV downstairs (via Ethernet!) and it really pollutes the results: https://imgur.com/a/F33LjeA
> >
> > I wouldn't say that pollutes the results, I'd say those are the
> >
> > typical and terrible results most users get when trying to do more
> >
> > than one thing at once on wifi or over the internet. Also it helps to
> >
> > have sqm on your isp updownlink, especially with netflix as it is
> >
> > bursty as heck
> >
> > I do not know if aql is enabled on that chipset. To my eye, it isn't,
> >
> > but the induced latency isn't quite high enough to represent what I
> >
> > viewed as the baseline problems.

Yup, AQL is enabled on this chipset, and I also have SQM piece_of_cake enabled to 50/50 mbps. I ran another test and made sure that there was no significant activity on the wireless network before I started it. Here are the results: https://imgur.com/a/7MzMVVJ

And the ping to 1.1.1.1 has the same shape: https://imgur.com/a/drcTGfv

Looks quite weird to me with all the oscillating latency spikes.
> >
> > https://forum.openwrt.org/t/aql-and-the-ath10k-is-lovely/59002
> >
> > Do a packet capture at the AP and look for multciast bursts especially

I think I'll have to find a moment where I have the network to myself again to be able to properly test things and figure out what might be causing this stuff. I'll probably do a packet capture at some point.

> >
> > And I keep trying to set aside time to go deep on AQL again, as the
> >
> > 7800 is a pretty popular box...
> >
> > > For what it's worth, the ping to 1.1.1.1 looks the same.
> > >
> > > Michael
> > >
> > > ‐‐‐‐‐‐‐ Original Message ‐‐‐‐‐‐‐
> > >
> > > On Friday, January 15th, 2021 at 17:23, Toke Høiland-Jørgensen toke at toke.dk wrote:
> > >
> > > > Michael Yartys via Make-wifi-fast make-wifi-fast at lists.bufferbloat.net
> > > >
> > > > writes:
> > > >
> > > > > Hi
> > > > >
> > > > > I decided to run a couple of flent 8-stream TCP downloads from
> > > > >
> > > > > netperf-eu.bufferbloat.net to my Fedora 33 laptop, and I noticed some
> > > > >
> > > > > weird periodic latency in the resulting graphs. My router is the
> > > > >
> > > > > NETGEAR R7800 running OpenWrt with an ath10k radio. Here's a graph
> > > > >
> > > > > that's representative of the results that I got:
> > > > >
> > > > > https://imgur.com/a/BOKNzht
> > > > >
> > > > > This looks similar to the latency spikes you would get from channel
> > > > >
> > > > > scans, but I have specified the BSSID in the Gnome network settings,
> > > > >
> > > > > which means that background scanning should be disabled. In addition,
> > > > >
> > > > > the test ran for 30 minutes, and channel scans are much more frequent
> > > > >
> > > > > than the latency spikes seen in the graph above. Here's what channel
> > > > >
> > > > > scans look like in a 15 minute long test: https://imgur.com/a/YJepfcp
> > > > >
> > > > > Does anyone here know what could be causing this behaviour?
> > > >
> > > > Hmm, your email reminds me that the server you are testing against was
> > > >
> > > > having some issue before the holidays. So it may just be that it's
> > > >
> > > > crapping out and it has nothing to do with your connection.
> > > >
> > > > You could try adding a secondary ping flow (--test-parameter
> > > >
> > > > ping_hosts=one.one.one.one should do it) and see if that secondary flow
> > > >
> > > > also shows the same spikes. If it does, it's likely your connection, if
> > > >
> > > > not it's likely the server...
> > > >
> > > > -Toke
> > >
> > > 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
>
> --
>
> "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