[Bloat] DOCSIS 3.1 support for AQM

[Bloat] DOCSIS 3.1 support for AQM

[Greg White]

[-- Attachment #1: Type: text/plain, Size: 575 bytes --]

FYI, the DOCSIS 3.1 specifications were released yesterday.   New in D3.1 is mandatory support for AQM (with default = on).

Cable modems are required to implement a variant of PIE (detailed in Annex M of MAC and Upper Layer Protocols Interface Specification<http://www.cablelabs.com/specifications/CM-SP-MULPIv3.1-101-131029.pdf>) as the default AQM.  Modem vendors are free to implement additional algorithms if they so choose.  CMTSs are required to implement an AQM meeting certain high-level criteria, but the choice of algorithm is left to the implementer.

-Greg

[-- Attachment #2: Type: text/html, Size: 955 bytes --]

Re: [Bloat] [aqm] DOCSIS 3.1 support for AQM

[Bless, Roland (TM)]

Hi Greg,

Am 30.10.2013 21:13, schrieb Greg White:
> Cable modems are required to implement a variant of PIE (detailed in
> Annex M of MAC and Upper Layer Protocols Interface Specification
> <http://www.cablelabs.com/specifications/CM-SP-MULPIv3.1-101-131029.pdf>) as
> the default AQM.  Modem vendors are free to implement additional
> algorithms if they so choose.  CMTSs are required to implement an AQM
> meeting certain high-level criteria, but the choice of algorithm is left
> to the implementer.

Thanks for the information. I'd be interested in why you have chosen
PIE, e.g., instead of sfq-CoDel. Any pointers to evaluation
reports/results? Last time I saw a presentation on this it seemed
that CoDel was performing quite well.

Regards,
 Roland

Re: [Bloat] [aqm] DOCSIS 3.1 support for AQM

[Aaron Wood]

[-- Attachment #1: Type: text/plain, Size: 907 bytes --]

> Thanks for the information. I'd be interested in why you have chosen
> PIE, e.g., instead of sfq-CoDel. Any pointers to evaluation
> reports/results? Last time I saw a presentation on this it seemed
> that CoDel was performing quite well.
>

I think this cablelabs report makes the argument for PIE:

http://www.cablelabs.com/downloads/pubs/Active_Queue_Management_Algorithms_DOCSIS_3_0.pdf

Mostly in that in the heavy traffic scenarios, PIE outperforms sfq_codel,
and in general is a tad bit better than codel, with a simpler
implementation (I think).  Although I think I take issue with the "heavy
traffic" model, but I'm guessing (hoping) that it's based on surveys of
customer traffic.  60-110 upstream flows seems like a lot.  But it's based
around a heavy use of BitTorrent, so maybe that's reasonable for some
people.

But in all other cases, sfq really blows the doors off of the others.

-Aaron

[-- Attachment #2: Type: text/html, Size: 1440 bytes --]

Re: [Bloat] [aqm] DOCSIS 3.1 support for AQM

[Michael Richardson]

Has any one any analysis of having sfq_codel on the CPE (uplink) stream,
and PIE on the CMTS (downlink) stream?

I'm assuming that the two directions should not conflict, but given that
RTT estimates and congestion signals involve bidirectional traffic, maybe
this assumption is wrong.

I also want to suggest that the fairness considerations of the household
are not necessarily the same as the fairness considerations of the ISP.

--
]               Never tell me the odds!                 | ipv6 mesh networks [
]   Michael Richardson, Sandelman Software Works        | network architect  [
]     mcr@sandelman.ca  http://www.sandelman.ca/        |   ruby on rails    [

Re: [Bloat] [aqm] DOCSIS 3.1 support for AQM

[Greg White]

[-- Attachment #1: Type: text/plain, Size: 1511 bytes --]

Rong will cover this at a high-level during the IETF AQM session tomorrow.  Her slides are posted:

?www.ietf.org/proceedings/88/slides/slides-88-aqm-1.pdf<http://www.ietf.org/proceedings/88/slides/slides-88-aqm-1.pdf>

I plan to do a follow-up to the paper you linked to below to give some of the details.  Should be ready before IETF89.

-Greg

From: Aaron Wood <woody77@gmail.com<mailto:woody77@gmail.com>>
Date: Thursday, October 31, 2013 at 7:23 AM
To: bloat <bloat@lists.bufferbloat.net<mailto:bloat@lists.bufferbloat.net>>
Subject: Re: [Bloat] [aqm] DOCSIS 3.1 support for AQM


Thanks for the information. I'd be interested in why you have chosen
PIE, e.g., instead of sfq-CoDel. Any pointers to evaluation
reports/results? Last time I saw a presentation on this it seemed
that CoDel was performing quite well.

I think this cablelabs report makes the argument for PIE:

http://www.cablelabs.com/downloads/pubs/Active_Queue_Management_Algorithms_DOCSIS_3_0.pdf

Mostly in that in the heavy traffic scenarios, PIE outperforms sfq_codel, and in general is a tad bit better than codel, with a simpler implementation (I think).  Although I think I take issue with the "heavy traffic" model, but I'm guessing (hoping) that it's based on surveys of customer traffic.  60-110 upstream flows seems like a lot.  But it's based around a heavy use of BitTorrent, so maybe that's reasonable for some people.

But in all other cases, sfq really blows the doors off of the others.

-Aaron

[-- Attachment #2: Type: text/html, Size: 3076 bytes --]