[Bloat] RE : I feel an urge to update this

[Bloat] RE : I feel an urge to update this

[luca.muscariello]

[-- Attachment #1: Type: text/plain, Size: 3571 bytes --]

Dave

I feel like you're rather optimistic about the deployment of fq
Also, you mention uplink in home gateways. There you not only have bittorrent but also a lot of personal cloud going up, a lot of video calls.

There IW10 without fq is going to be bad. In the US your are going to have pie on cable if I am not wrong...uhm...

Luca



-------- Message d'origine --------
De : Dave Taht <dave.taht@gmail.com>
Date : 2014/09/25 15:26 (GMT+01:00)
A : Mikael Abrahamsson <swmike@swm.pp.se>
Cc : bloat <bloat@lists.bufferbloat.net>
Objet : Re: [Bloat] I feel an urge to update this


On Thu, Sep 25, 2014 at 6:00 AM, Mikael Abrahamsson <swmike@swm.pp.se> wrote:
> On Thu, 25 Sep 2014, David Lang wrote:
>
>> What is the problem with making this assumption? Why should we try to
>> change every device on the Internet to provide this information instead of
>> just using this as the default?
>
> Read the email that started this thread.

One saving grace of the IW10 deployment so far is that it is (mostly)
limited to Linux, and that methods, such as pacing/initial
spreading/fq, to cope with it better, are developing concurrently.

It's not all bad - on GigE ethernet networks, it's ok, IW10 may well
be useful in wifi packet aggregation, etc, but it does do potentially
a great deal of damage to networks running at very low rates. My
concern has largely been the collateral damage it causes network
uploads from the edge, where vastly lower rates are common.

It could be mitigated further in a desktop deployment by (for example)
sticking with IW4 for everything going out a default gateway, and
steps taken to make sure applications like bittorrent/transmission
used a lower default. I have no idea what the IW is for android, but
that should perhaps be lower also. I think ledbat shouldn't have an IW
at all....

I would not mind if IW information was carried as a payload in dhcp/hnetd/etc...

I have no problem with people suggesting changes here that require
"changing the entire internet". We seem to have done that a couple
times, already. :) It would be nice if we could calculate the damage
it causes along whatever portion of the internet is running at below
10Mbit....

A simple question - how many T1 lines are left in the world?



> --
> Mikael Abrahamsson    email: swmike@swm.pp.se



--
Dave Täht

https://www.bufferbloat.net/projects/make-wifi-fast
_______________________________________________
Bloat mailing list
Bloat@lists.bufferbloat.net
https://lists.bufferbloat.net/listinfo/bloat

_________________________________________________________________________________________________________________________

Ce message et ses pieces jointes peuvent contenir des informations confidentielles ou privilegiees et ne doivent donc
pas etre diffuses, exploites ou copies sans autorisation. Si vous avez recu ce message par erreur, veuillez le signaler
a l'expediteur et le detruire ainsi que les pieces jointes. Les messages electroniques etant susceptibles d'alteration,
Orange decline toute responsabilite si ce message a ete altere, deforme ou falsifie. Merci.

This message and its attachments may contain confidential or privileged information that may be protected by law;
they should not be distributed, used or copied without authorisation.
If you have received this email in error, please notify the sender and delete this message and its attachments.
As emails may be altered, Orange is not liable for messages that have been modified, changed or falsified.
Thank you.


[-- Attachment #2: Type: text/html, Size: 4395 bytes --]

Re: [Bloat] RE : I feel an urge to update this

[Dave Taht]

On Sat, Sep 27, 2014 at 11:07 AM,  <luca.muscariello@orange.com> wrote:
> Dave
>
> I feel like you're rather optimistic about the deployment of fq

Hmm? At this point I'm rather sanguine about it. Every new high end
home router product
entering the market I've tried has it in it under some brand name or
another. Admittedly
my sample set is biased by anyone advertising better QoS...

Take for example netgear's new "nighthawk x4" products' "Dynamic QoS" feature.

http://snapon.lab.bufferbloat.net/~d/netgear_x4/netgear_x4_22_55_dynamic_qos.png

This is an enormous improvement on their earlier QoS systems, which
weren't deserving of the name.

While I don't know for sure that it's fq_codel underneath, it sure
looks like it, (from the interleave of the captures also) with an
optimization for ping. (I disagree with that, preferring to
de-optimize ping,  but it does make benchmarks look good. ECN is
disabled, sadly enough, and they don't offer options for DSL framing
compensation...)

The rrul curves for AQM would look very different, and sfq alone would
have larger latency spikes on the tcp flows.

With netgear, dd-link, buffalo, openwrt barrier breaker and everything
derived from it, dd-wrt, ipfire, pfsense, several others I forget, so
far on board, that's a pretty clean sweep of the non-isp-based home
router market, both with shipping product and aftermarket firmware. On
the higher end ubnt is making progress (the enthusiasm over there has
to be seen to be believed), mikrotik has shown a bit of interest, and
I think some more of the mid-range "cloudy" enterprise wifi makers are
paying attention. I know of several ISPs that paid attention and are
shipping fq_codel based product, also.

Of course there are some problems on the lower end boxes in getting
htb to scale, I hope we have some answers for that upcoming. And
there's work to be done on many a DSL device for BQL support, and not
a lot a sign of life from the head end makers except for tantalizing
hints from arris, and there are still issues with doing better
classification that would be nice to resolve globally.

So we are looking at deployments measured in the millions in the
coming year, easily.

> Also, you mention uplink in home gateways. There you not only have
> bittorrent but also a lot of personal cloud going up, a lot of video calls.

Yes, torrent is a toughie, but I take comfort in that most torrent
users are only doing 6 reno flows at a time, and if folk use a three
level system like sqm and mark the torrent traffic CS1 (or have DPI as
it appears the brand names are doing),
we're done there, too. Web traffic (being almost entirely slow start)
cuts through torrent traffic like butter, anyway.

The jury on videoconferencing on the interactions of fq, aqm, or
fq+aqm is still out. Most of the benchmarks I've seen have been taken
at very low rates (1-4mbit, where I'm far more interesting in rates
like 16/3 and higher), but that is looking good also, in all cases an
improvement over drop tail. I would like it if they were protecting
the main frame with ecn, at least....

>
> There IW10 without fq is going to be bad. In the US your are going to have
> pie on cable if I am not wrong...uhm...

Well, I hear DOCSIS 3.1 trials w/pie will start mid-next year. If
people don't dig pie, they can always slam one of these fancy new
boxes in front.

But I still dislike IW10 being on by default on boxes not based in the
data center.

> A simple question - how many T1 lines are left in the world?

It really was a simple question, I figured it could have been answered by
inspecting a few regulatory databases somewhere....

>
>> --
>> Mikael Abrahamsson    email: swmike@swm.pp.se
>
>
>
> --
> Dave Täht
>
> https://www.bufferbloat.net/projects/make-wifi-fast
> _______________________________________________
> Bloat mailing list
> Bloat@lists.bufferbloat.net
> https://lists.bufferbloat.net/listinfo/bloat
>
> _________________________________________________________________________________________________________________________
>
> Ce message et ses pieces jointes peuvent contenir des informations
> confidentielles ou privilegiees et ne doivent donc
> pas etre diffuses, exploites ou copies sans autorisation. Si vous avez recu
> ce message par erreur, veuillez le signaler
> a l'expediteur et le detruire ainsi que les pieces jointes. Les messages
> electroniques etant susceptibles d'alteration,
> Orange decline toute responsabilite si ce message a ete altere, deforme ou
> falsifie. Merci.
>
> This message and its attachments may contain confidential or privileged
> information that may be protected by law;
> they should not be distributed, used or copied without authorisation.
> If you have received this email in error, please notify the sender and
> delete this message and its attachments.
> As emails may be altered, Orange is not liable for messages that have been
> modified, changed or falsified.
> Thank you.



-- 
Dave Täht

https://www.bufferbloat.net/projects/make-wifi-fast

Re: [Bloat] RE : I feel an urge to update this

[Dave Taht]

I guess I should try to be less unhappy about IW10 by default. It
could be worse.

http://www.cdnplanet.com/blog/initcwnd-settings-major-cdn-providers/

-- 
Dave Täht

https://www.bufferbloat.net/projects/make-wifi-fast

Re: [Bloat] RE : I feel an urge to update this

[Rick Jones]

On 09/27/2014 02:57 PM, Dave Taht wrote:
> I guess I should try to be less unhappy about IW10 by default. It
> could be worse.
>
> http://www.cdnplanet.com/blog/initcwnd-settings-major-cdn-providers/

I'm confused slightly about just when that data was sampled.  The top of 
the page claims 2014, but the age of several of the disqus comments is 
given as 3 years.

rick jones