[Make-wifi-fast] Questions from the Windows world, IQrouter, etc

[Aaron Wood]

This looks like it's using the same approach as the sqm_scripts (which does
rate-limiting and uses fq_codel), based on an initial sampling of the
interface rate, and the periodic measurements which are used to modify the
sqm settings.  I would be surprised if it didn't do better than most stock
routers.

What this doesn't do is make wifi faster/better.  It makes the connection
to the ISP behave better, and if that's the bottleneck that you have, then
that's enough.  But uploads where wifi is the bottleneck aren't going to
get much better.  The adapter/drivers make a big difference (radio quality,
antenna quality, driver buffer depth/management).

By default, most routers seem to be using a fixed size buffer in terms of
the number of packets, and somehow that ended up at 1000 packets.  That
means that if wifi is the bottleneck, there's:

~120ms of latency in the buffer at 100Mbps
~250ms of latency at 50Mbps
~0.5 sec of latency at 25 Mbps
~1.2 sec at 10Mbps
etc.

-Aaron

On Mon, Sep 12, 2016 at 10:47 AM, Jon Pike <jonpike54 at gmail.com> wrote:

> Have been watching and appriciating all the progress going on...  wish I
> had a better understanding and better coding skills to contribute somehow.
>
> Don't know if this is the best place to bring it up,  I certainly don't
> want to get in the way of Real Work going on...  but on my last few visits
> to DSLreports  Speed test page, I've seen an ad for the IQrouter...
> http://evenroute.com/
>
> This new?  Any connection with FQ-Codel,  etc?  By the definition, sounds
> like they may be rolling their own,  with the dynamic measuring and
> adjusting.  At the very least it's a prominent and pretty well described
> mention of Bufferbloat problems and what can fix them.  Apologies if this
> is old news, but I hadn't seen them or the ad before.  It's also
> interesting in that I've been looking for a good router upgrade,  and
> looking at TP-LINK C5 and C7s...  and the IQrouter looks exactly like a
> C5,  except for the N900 on the right edge. (rather than AC1200)  Must have
> lucked into a huge warehouse of early models?
>
> A noobish question I have,  is how much influence does the client side of
> a network have on the overall link quality?  Have been setting up a few
> wifi connections in the house,  and have noticed drastic differences in the
> amount of bloat between two PC's.  Two different USB dongles,  and of
> course two different positions.  Have yet to swap the two adapters, which
> would make it a fair test.  But I wondered how much influence the
> adapter/driver has on overall flow?  And yes,  these are Windows machines,
> which brings up another question, how much difference might there be in a
> Linux/Win setup, vs Linux/Linux?  Or, is that a territory that needs to be
> explored?
>
> In that case,  perhaps someone can point me to an equivalent to "FLENT for
> Dummies".
>
> _______________________________________________
> Make-wifi-fast mailing list
> Make-wifi-fast at lists.bufferbloat.net
> https://lists.bufferbloat.net/listinfo/make-wifi-fast
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.bufferbloat.net/pipermail/make-wifi-fast/attachments/20160912/a07d0c3b/attachment.html>