[Nottingham] The arguments with WiFi continue

[Michael Quaintance]

Jason, what I think you're experiencing here is traffic shaping and
"marketing bulls..t". It seems like the same sort of crappiness I get from
Comcast over here in Silicon Valley.

Comcast deliberately prioritises traffic to speedtest.net, and presumably
to fast.com, but I've not tried that yet, whereas to my Linode servers
where I have minimum of gigabit connections, I don't get anything like the
sort of speed I pay for (100Mbps), and if I go through OpenVPN, these get
throttled to <10Mbps. This way, it looks like you are getting your full
connection speed even if practically, you're being held back everywhere
else.

Unfortunately, as a single consumer, especially on a home connection rather
than a business one, there is almost nothing we can do about this.

If you're specifically trying to fix wifi performance issues, like I
thought in the beginning of this thread, you need to do iperf to machines
wired on your local LAN so that nefarious traffic shaping by your ISP is
out of the picture.

If you're trying to fix performance to cloud servers you operate, try to
stick to well-known ports that are less likely to be traffic shaped...
80/443:tcp (http/https) are generally good, 22:tcp (ssh) is often quite
decent as well, and in my experience 1194:udp (openvpn) is terrible. But
the port number alone may not be enough, generally ISPs use deep packet
inspection to determine the type of traffic in order to apply traffic
shaping.

Because of this DPI, my next step is to try running my OpenVPN connections
through STunnel, so they really look like HTTPS traffic in the hope that is
throttled less aggressively.

Good luck!

-Michael


On Thu, May 19, 2016 at 4:27 AM, Martin via Nottingham <
nottingham at mailman.lug.org.uk> wrote:

> On 19/05/16 09:31, Jason Irwin via Nottingham wrote:
> > On 15/05/16 19:16, Jason Irwin wrote:
> >> Minor addendum:
> > Further addendum, and this gets weird.
> >
> > Ethernet: ~950Mbps, seems legit, it's gigabit throughout the house.
> > 5Ghz: ~30Mbps, OK...that's slow.
> > 2.4Ghz: ~19Mbps, that's slower....
> >
> > Now, for the weirdness.
> > First off, I access fast.com over ethernet at get 150Mbps. That's
> > expected, that's the speed of my downlink and it proves I'm not hitting
> > some limit.
> > I access it again over 5GHz from the exact same machine where I just ran
> > all the above tests; what speed do I get reported? ~30MBps?
> > No, 70+MBps. WTF?
> >
> > I would blame DD-WRT but I have used two separate routers, one of which
> > isn't DD-WRT, to check my LAN speed and both give the same results.
> > 70Mbps still isn't great, but it's twice as fast as the test to my own
> > server.
> > That make no sense at all, unless "iperf" is giving wonky results or my
> > comprehension of networking is even worse than I thought it was!
>
> All very plausible...
>
> Are you missing the effect of sometimes/sometimes-not aggregating the
> multiple bandwidths from MIMO and multiple *band* operation?...
>
> This WiFi stuff can get a little convoluted ;-)
>
>
> Good luck for sleuthing...
>
> Cheers,
> Martin
>
>
>
> --
> - ╔═══════════════════╦══════════════════════════════════════════╗
> - ║   Martin Lomas    ║ OpenPGP (GPG/PGP) Public Key: 0xCEE1D3B7 ║
> - ║ martin@ ml1 co uk ║ Import from   hkp://subkeys.pgp.net   or ║
> - ║ ----------------- ║ http:// ml1 .co .uk/martin_ml1_co_uk.gpg ║
> - ╚═══════════════════╩══════════════════════════════════════════╝
>
> _______________________________________________
> Nottingham mailing list
> Nottingham at mailman.lug.org.uk
> https://mailman.lug.org.uk/mailman/listinfo/nottingham
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.lug.org.uk/pipermail/nottingham/attachments/20160519/9ade8852/attachment.html>