[Bloat] Up-to-date buffer sizes?

[Bloat] Up-to-date buffer sizes?

[Michael Menth]

Hi all,

are there up-to-date references giving evidence about typical buffer 
sizes for various link speeds and technologies?

Kind regards

Michael

-- 
Prof. Dr. habil. Michael Menth
University of Tuebingen
Faculty of Science
Department of Computer Science
Chair of Communication Networks
Sand 13, 72076 Tuebingen, Germany
phone: (+49)-7071/29-70505
fax: (+49)-7071/29-5220
mailto:menth@uni-tuebingen.de
http://kn.inf.uni-tuebingen.de

Re: [Bloat] Up-to-date buffer sizes?

[Toke Høiland-Jørgensen]

Michael Menth <menth@uni-tuebingen.de> writes:

> Hi all,
>
> are there up-to-date references giving evidence about typical buffer 
> sizes for various link speeds and technologies?

Heh. There was a whole workshop on it a couple of years ago; not sure if
it concluded anything: http://buffer-workshop.stanford.edu/program/

But really, asking about buffer sizing is missing the point; if you have
static buffers with no other management (like AQM and FQ) you're most
likely already doing it wrong... :)

-Toke

Re: [Bloat] Up-to-date buffer sizes?

[Jesper Dangaard Brouer]

On 09/03/2022 17.31, Toke Høiland-Jørgensen via Bloat wrote:
> Michael Menth <menth@uni-tuebingen.de> writes:
> 
>> Hi all,
>>
>> are there up-to-date references giving evidence about typical buffer
>> sizes for various link speeds and technologies?
> 
> Heh. There was a whole workshop on it a couple of years ago; not sure if
> it concluded anything: http://buffer-workshop.stanford.edu/program/
> 
> But really, asking about buffer sizing is missing the point; if you have
> static buffers with no other management (like AQM and FQ) you're most
> likely already doing it wrong... :)

Exactly, I agree with Toke. The important parameter is the latency.
Or the packet sojourn time (rfc8289 + rfc8290) observed waiting in the 
queue.

The question you should be asking is:
  - What is the max queue latency I'm "willing" to experience on this link?

Hint, you can then depending on the link rate calculate the max buffer 
size you should configure.

The short solution is:
  - just use fq_codel (rfc8290) as the default qdisc.

--Jesper

Re: [Bloat] Up-to-date buffer sizes?

[Michael Menth]

Hi all,

I don't question the usefulness of AQMs for buffers - on the contrary. 
But what are up-to-date buffer sizes in networking gears, especially if 
AQMs are not in use? It's hard to find public and information about it. 
Anyone can point to a citable source?

This raises also the question about the deployment of AQMs in networking 
infrastructure. I know it's already adopted by some OSs, but what about 
forwarding nodes? Any papers about it?

Kind regards

Michael

Am 09.03.2022 um 18:24 schrieb Jesper Dangaard Brouer:
> 
> 
> On 09/03/2022 17.31, Toke Høiland-Jørgensen via Bloat wrote:
>> Michael Menth <menth@uni-tuebingen.de> writes:
>>
>>> Hi all,
>>>
>>> are there up-to-date references giving evidence about typical buffer
>>> sizes for various link speeds and technologies?
>>
>> Heh. There was a whole workshop on it a couple of years ago; not sure if
>> it concluded anything: http://buffer-workshop.stanford.edu/program/
>>
>> But really, asking about buffer sizing is missing the point; if you have
>> static buffers with no other management (like AQM and FQ) you're most
>> likely already doing it wrong... :)
> 
> Exactly, I agree with Toke. The important parameter is the latency.
> Or the packet sojourn time (rfc8289 + rfc8290) observed waiting in the 
> queue.
> 
> The question you should be asking is:
>   - What is the max queue latency I'm "willing" to experience on this link?
> 
> Hint, you can then depending on the link rate calculate the max buffer 
> size you should configure.
> 
> The short solution is:
>   - just use fq_codel (rfc8290) as the default qdisc.
> 
> --Jesper
> 
> 
> 

-- 
Prof. Dr. habil. Michael Menth
University of Tuebingen
Faculty of Science
Department of Computer Science
Chair of Communication Networks
Sand 13, 72076 Tuebingen, Germany
phone: (+49)-7071/29-70505
fax: (+49)-7071/29-5220
mailto:menth@uni-tuebingen.de
http://kn.inf.uni-tuebingen.de

Re: [Bloat] Up-to-date buffer sizes?

[Aaron Wood]

[-- Attachment #1: Type: text/plain, Size: 2449 bytes --]

Are you asking what they _should_ be, or what the typical buffering seen in
equipment actually is?


On Wed, Mar 9, 2022 at 9:39 AM Michael Menth <menth@uni-tuebingen.de> wrote:

> Hi all,
>
> I don't question the usefulness of AQMs for buffers - on the contrary.
> But what are up-to-date buffer sizes in networking gears, especially if
> AQMs are not in use? It's hard to find public and information about it.
> Anyone can point to a citable source?
>
> This raises also the question about the deployment of AQMs in networking
> infrastructure. I know it's already adopted by some OSs, but what about
> forwarding nodes? Any papers about it?
>
> Kind regards
>
> Michael
>
> Am 09.03.2022 um 18:24 schrieb Jesper Dangaard Brouer:
> >
> >
> > On 09/03/2022 17.31, Toke Høiland-Jørgensen via Bloat wrote:
> >> Michael Menth <menth@uni-tuebingen.de> writes:
> >>
> >>> Hi all,
> >>>
> >>> are there up-to-date references giving evidence about typical buffer
> >>> sizes for various link speeds and technologies?
> >>
> >> Heh. There was a whole workshop on it a couple of years ago; not sure if
> >> it concluded anything: http://buffer-workshop.stanford.edu/program/
> >>
> >> But really, asking about buffer sizing is missing the point; if you have
> >> static buffers with no other management (like AQM and FQ) you're most
> >> likely already doing it wrong... :)
> >
> > Exactly, I agree with Toke. The important parameter is the latency.
> > Or the packet sojourn time (rfc8289 + rfc8290) observed waiting in the
> > queue.
> >
> > The question you should be asking is:
> >   - What is the max queue latency I'm "willing" to experience on this
> link?
> >
> > Hint, you can then depending on the link rate calculate the max buffer
> > size you should configure.
> >
> > The short solution is:
> >   - just use fq_codel (rfc8290) as the default qdisc.
> >
> > --Jesper
> >
> >
> >
>
> --
> Prof. Dr. habil. Michael Menth
> University of Tuebingen
> Faculty of Science
> Department of Computer Science
> Chair of Communication Networks
> Sand 13, 72076 Tuebingen, Germany
> phone: (+49)-7071/29-70505
> fax: (+49)-7071/29-5220
> mailto:menth@uni-tuebingen.de
> http://kn.inf.uni-tuebingen.de
> _______________________________________________
> Bloat mailing list
> Bloat@lists.bufferbloat.net
> https://lists.bufferbloat.net/listinfo/bloat
>
-- 
- Sent from my iPhone.

[-- Attachment #2: Type: text/html, Size: 3611 bytes --]

Re: [Bloat] Up-to-date buffer sizes?

[David Lang]

[-- Attachment #1: Type: text/plain, Size: 2303 bytes --]

If the link is not a bottleneck, then you will not be using buffers (whatever 
they are configured to be)

networking gear tends toward the proprietary (although there's a growing amount 
that's linux based now) and they tend to be very closed mouth about 
configuration like this.

David Lang

  On Wed, 9 Mar 2022, Michael Menth wrote:

> Date: Wed, 9 Mar 2022 18:39:08 +0100
> From: Michael Menth <menth@uni-tuebingen.de>
> To: Jesper Dangaard Brouer <jbrouer@redhat.com>,
>     Toke Høiland-Jørgensen <toke@toke.dk>, bloat <bloat@lists.bufferbloat.net>
> Subject: Re: [Bloat] Up-to-date buffer sizes?
> 
> Hi all,
>
> I don't question the usefulness of AQMs for buffers - on the contrary. 
> But what are up-to-date buffer sizes in networking gears, especially if 
> AQMs are not in use? It's hard to find public and information about it. 
> Anyone can point to a citable source?
>
> This raises also the question about the deployment of AQMs in networking 
> infrastructure. I know it's already adopted by some OSs, but what about 
> forwarding nodes? Any papers about it?
>
> Kind regards
>
> Michael
>
> Am 09.03.2022 um 18:24 schrieb Jesper Dangaard Brouer:
>> 
>> 
>> On 09/03/2022 17.31, Toke Høiland-Jørgensen via Bloat wrote:
>>> Michael Menth <menth@uni-tuebingen.de> writes:
>>>
>>>> Hi all,
>>>>
>>>> are there up-to-date references giving evidence about typical buffer
>>>> sizes for various link speeds and technologies?
>>>
>>> Heh. There was a whole workshop on it a couple of years ago; not sure if
>>> it concluded anything: http://buffer-workshop.stanford.edu/program/
>>>
>>> But really, asking about buffer sizing is missing the point; if you have
>>> static buffers with no other management (like AQM and FQ) you're most
>>> likely already doing it wrong... :)
>> 
>> Exactly, I agree with Toke. The important parameter is the latency.
>> Or the packet sojourn time (rfc8289 + rfc8290) observed waiting in the 
>> queue.
>> 
>> The question you should be asking is:
>>   - What is the max queue latency I'm "willing" to experience on this link?
>> 
>> Hint, you can then depending on the link rate calculate the max buffer 
>> size you should configure.
>> 
>> The short solution is:
>>   - just use fq_codel (rfc8290) as the default qdisc.
>> 
>> --Jesper
>> 
>> 
>> 
>
>

Re: [Bloat] Up-to-date buffer sizes?

[Amr Rizk]

These are some recent Telco measurements (Figure 3) from one of the papers from that workshop

http://buffer-workshop.stanford.edu/papers/paper7.pdf

There on can observe that most flows in Internet are <60 ms. I assume applying Appenzeller's rule is not a bad idea for static buffer sizing. However, as Toke said, AQM is of course better..

Best
Amr

-----Ursprüngliche Nachricht-----
Von: Bloat <bloat-bounces@lists.bufferbloat.net> Im Auftrag von Toke Høiland-Jørgensen via Bloat
Gesendet: Mittwoch, 9. März 2022 17:32
An: Michael Menth <menth@uni-tuebingen.de>; bloat <bloat@lists.bufferbloat.net>
Betreff: Re: [Bloat] Up-to-date buffer sizes?

Michael Menth <menth@uni-tuebingen.de> writes:

> Hi all,
>
> are there up-to-date references giving evidence about typical buffer 
> sizes for various link speeds and technologies?

Heh. There was a whole workshop on it a couple of years ago; not sure if it concluded anything: http://buffer-workshop.stanford.edu/program/

But really, asking about buffer sizing is missing the point; if you have static buffers with no other management (like AQM and FQ) you're most likely already doing it wrong... :)

-Toke
_______________________________________________
Bloat mailing list
Bloat@lists.bufferbloat.net
https://lists.bufferbloat.net/listinfo/bloat

Re: [Bloat] Up-to-date buffer sizes?

[Jonas Mårtensson]

[-- Attachment #1: Type: text/plain, Size: 2498 bytes --]

Here is a pretty good list but it's focused on data center and carrier
networking gear, not so much home networking gear:

https://people.ucsc.edu/~warner/buffer.html

/Jonas

On Wed, Mar 9, 2022 at 6:39 PM Michael Menth <menth@uni-tuebingen.de> wrote:

> Hi all,
>
> I don't question the usefulness of AQMs for buffers - on the contrary.
> But what are up-to-date buffer sizes in networking gears, especially if
> AQMs are not in use? It's hard to find public and information about it.
> Anyone can point to a citable source?
>
> This raises also the question about the deployment of AQMs in networking
> infrastructure. I know it's already adopted by some OSs, but what about
> forwarding nodes? Any papers about it?
>
> Kind regards
>
> Michael
>
> Am 09.03.2022 um 18:24 schrieb Jesper Dangaard Brouer:
> >
> >
> > On 09/03/2022 17.31, Toke Høiland-Jørgensen via Bloat wrote:
> >> Michael Menth <menth@uni-tuebingen.de> writes:
> >>
> >>> Hi all,
> >>>
> >>> are there up-to-date references giving evidence about typical buffer
> >>> sizes for various link speeds and technologies?
> >>
> >> Heh. There was a whole workshop on it a couple of years ago; not sure if
> >> it concluded anything: http://buffer-workshop.stanford.edu/program/
> >>
> >> But really, asking about buffer sizing is missing the point; if you have
> >> static buffers with no other management (like AQM and FQ) you're most
> >> likely already doing it wrong... :)
> >
> > Exactly, I agree with Toke. The important parameter is the latency.
> > Or the packet sojourn time (rfc8289 + rfc8290) observed waiting in the
> > queue.
> >
> > The question you should be asking is:
> >   - What is the max queue latency I'm "willing" to experience on this
> link?
> >
> > Hint, you can then depending on the link rate calculate the max buffer
> > size you should configure.
> >
> > The short solution is:
> >   - just use fq_codel (rfc8290) as the default qdisc.
> >
> > --Jesper
> >
> >
> >
>
> --
> Prof. Dr. habil. Michael Menth
> University of Tuebingen
> Faculty of Science
> Department of Computer Science
> Chair of Communication Networks
> Sand 13, 72076 Tuebingen, Germany
> phone: (+49)-7071/29-70505
> fax: (+49)-7071/29-5220
> mailto:menth@uni-tuebingen.de
> http://kn.inf.uni-tuebingen.de
> _______________________________________________
> Bloat mailing list
> Bloat@lists.bufferbloat.net
> https://lists.bufferbloat.net/listinfo/bloat
>

[-- Attachment #2: Type: text/html, Size: 3675 bytes --]