Re: [tsvwg] feedback and thoughts L4S / SCE

[Mikael Abrahamsson <[e-mail]>]

On Sun, 22 Nov 2020, Sebastian Moeller wrote:

> just as a data point, OpenWrt switched to fq_codel as default qdisc on all interfaces some years ago.

And how often is this fq_codel the slowest link in the chain?

> I have no numbers how many OpenWrt devices are in the wild, and how many 
> commercial home router OSes are based on sufficiently modern OpenWrt, 
> but that is some use. The OpenWrt project also has no official counter

Most of the commercial openwrt based devices will not use CPU based 
forwarding (required to make fq_codel work) but instead has hardware 
accelerator (which does not do fq_codel).

> of running instances but for example Qualcomm's QCA Software Development 
> Kit (QSDK) is based on OpenWrt, so OpenWrt is more pervasive than one 
> would naively assume. That said, the QDSK kernel seems to be quite old 
> (at least Kernel major version 2 seems dead now), and I have no insight 
> which features are back-ported to it.

If you're using QCA hw-accelerated forwarding then you're not doing 
fq_codel.

> 	[SM] Again a prediction, not a fact, as I said reasonably recent 
> OpenWrt defaults to fq_codel on all interfaces (that support it) as well 
> as on WiFi which with inceasing access rates more and more often becomes 
> the true bottleneck (no numbers, as that is my prediction, based on my 
> own experience).

If you're using HW acceleration, then fq_codel isn't used also on wifi.

> 	[SM] Given our lack of hard numbers that is telling more about your  judgement than about reality.

Do you have numbers? How many ISP provided routers out there actually use 
fq_* ? ISP provided router is the most common in use.

> 	[SM] And that is pretty harsh. Also unhelpful teminology, "irrelevant" I would accept, but invalid is simply the wrong quality to argue about.

Fine.

> 	[SM] And here we are again at our lack of numbers... As before I share your intuition, but it is not a fact.
> "Fun" fact TCP Prague really is not at all suited to a FIFO world at all, as Pete demonstrated.

Do you have better numbers?

>> I also doubt we'll see widely deployed FQ going forward because of the cost of doing this in hw (and most devices are hw accelerated).
>
> 	[SM] An often heard claim that suffers from the fact that that has 
> simply never been tried at scale, so the cost argument reflects the 
> status quo not necessarily actual economic "laws", no?

Do you know of a FQ_* implementation in hardware?

> 	Also "wide-spread" use is again a claim that should be based on numbers (which cuts both ways, so applies to Pete's claim as well).

Thanks for ack:ing that.

> QUESTION: Where do you think most choke points live, at the ISP-end-user 
> boundary, deeper in the ISP's network, or even at exchange points? My 
> own limited experience points to the ISP-end-user edge, but that is 
> purely anecdotal.

The choke point is most often the customer last-mile access line or the 
wifi hop.

-- 
Mikael Abrahamsson    email: [e-mail]