1 edit
11 recommendations |
to EnasYorl
Re: [DOCSIS] MB8600 3.1 rocking during peak hourssaid by EnasYorl:... But I also am get getting the benefit of AQM that comes with a D3.1 modem and that is making a difference as well. I don't know if Comcast has implemented Active Queue Management (AQM) in any of their CMTSs yet, and if so, probably in just a few trial systems right now: From: » Re: [Plan] XI Gigabit DOCSIS 3.1 - data cap with 3-year agreement?said by jlivingood:While AQM is in the D3.1 specs it is not yet implemented. When it's ready from the D3.1 vendors we'll start a trial to evaluate it. TBD timing on that (at least on the DS, which is implemented in the CMTS, whereas US is implemented in the CPE). EDIT: My contacts at Comcast have confirmed that AQM has not yet been implemented in their network. |
|
|
1 recommendation |
AQM may not be implemented on the CMTS (ie: on downloads), but it's almost certainly enabled on all 3.1 modems (ie: on uploads).
AQM is important in both directions, but it's much more important on the upstream side because of how low 35 Mbps is in today's world.. |
|
|
EnasYorlThieves World join:2001-12-02 West
1 recommendation |
to telcodad
» cpham.perso.univ-pau.fr/ ··· CSIS.pdfDOCSIS-PIE Is required for 3.1 modems from my understanding. |
|
EnasYorl
1 recommendation |
7.6.2.2 CMTS AQM Requirements The CMTS MUST support a default AQM scheme defined by the vendor. The CMTS SHOULD support a published AQM algorithm as the default AQM. An AQM algorithm description that is publicly accessible allows for wider evaluation by the industry and networking community. The CMTS default AQM scheme MUST bound the median downstream packet forwarding latency in individual Service Flows. The CMTS default AQM scheme SHOULD allow each downstream Service Flow to attain and maintain a steady transfer rate at the Peak Traffic Rate before the Maximum Traffic Burst has been used. The CMTS default AQM scheme SHOULD allow each downstream Service Flow to attain and maintain a steady transfer rate at the Maximum Sustained Traffic Rate after the Maximum Traffic Burst has been used. The CMTS default AQM scheme MUST NOT use packet payload information which could identify the applications which are using the Service Flows. The CMTS default AQM scheme MUST work without manual tuning by the operator. The CMTS MUST support a configurable mechanism to control aspects of the AQM algorithm that affect trade-offs with other QoS requirements. The CMTS MUST be able to control AQM on a per-Service-Flow basis, including the ability to disable AQM. The CMTS default AQM scheme SHOULD comply with [RFC 7567]. The CMTS SHOULD minimize the number of buffered packets during the transition from Peak Traffic Rate to Maximum Sustained Traffic Rate. The CMTS SHOULD bound packet loss to an acceptable level for each of the Service Flows. The CMTS SHOULD adequately handle a variety of congestion avoidance methods that may be in use by transports and applications, such as TCP-Reno, TCP-CUBIC, TCP-SACK, LEDBAT, and RMCAT. The CMTS SHOULD disable or otherwise reset the AQM scheme for CMs operating in DOCSIS Light Sleep Mode. » www.scte.org/SCTEDocs/St ··· rds/SCTE 220-2 2016.pdf page 265 |
|
gtb Premium Member join:2016-05-16 united state
1 recommendation |
gtb
Premium Member
2018-Mar-24 10:36 am
said by EnasYorl:7.6.2.2 CMTS AQM Requirements And none of those require the operator to not disable it, only that it must be available (and no known provider in the US (outside of the possible test/stunt markets) have it enabled (and Comcast (according to telcodad) say's they do not)). AQM does have some benefits, and as with all other capabilities, it does have some drawbacks. |
|
3 recommendations |
said by gtb:said by EnasYorl:7.6.2.2 CMTS AQM Requirements And none of those require the operator to not disable it, only that it must be available (and no known provider in the US (outside of the possible test/stunt markets) have it enabled (and Comcast (according to telcodad) say's they do not)). Yet, Jason had said: said by jlivingood:While AQM is in the D3.1 specs it is not yet implemented. When it's ready from the D3.1 vendors we'll start a trial to evaluate it. ... So, if AQM is a D3.1 "requirement" how did those CMTSs/plugins end up receiving D3.1 certification from CableLabs? |
|
2 recommendations |
If you read closely: The CMTS MUST ****support**** a default AQM scheme defined by the vendor.
It's entirely possible that the CMTS supports it, and yet it is disabled for some reason. |
|
EnasYorlThieves World join:2001-12-02 West
2 recommendations |
to telcodad
Waiver with the promise to enable it later? |
|
5 recommendations |
to winterfl
said by winterfl:If you read closely: The CMTS MUST ****support**** a default AQM scheme defined by the vendor.
It's entirely possible that the CMTS supports it, and yet it is disabled for some reason. Sorry, but that still does not square with Jason's wording - "while AQM is in the D3.1 specs it is not yet implemented. When it's ready from the D3.1 vendors we'll start a trial to evaluate it. Not "implemented" and not "ready" implies more than just "disabled." |
|
EnasYorlThieves World join:2001-12-02 West |
CMTS or CPE? OR Both? |
|
1 recommendation |
I believe Jason was talking about AQM on the downstream, i.e., in the CMTS: said by jlivingood:While AQM is in the D3.1 specs it is not yet implemented. When it's ready from the D3.1 vendors we'll start a trial to evaluate it. TBD timing on that (at least on the DS, which is implemented in the CMTS, whereas US is implemented in the CPE). |
|
EnasYorlThieves World join:2001-12-02 West ·Ziply Fiber Motorola MB8600 Netgear RAX15 Asus RT-AC66
2 recommendations |
|
|
3 recommendations |
Thanks, that SCTE presentation from Viavi is excellent! |
|
gtb Premium Member join:2016-05-16 united state
1 recommendation |
to telcodad
said by telcodad:So, if AQM is a D3.1 "requirement" how did those CMTSs/plugins end up receiving D3.1 certification from CableLabs? CableLabs certifies CPE. Some vendors of CMTSs have "always" supported various QoS (which results in QoE) configurations, but depending on the specific firmware releases it may not support all the variants that are in the entire AQM portfolio (and it is not required to support all in any case), and some of those long known QoS variants can require a *lot* of careful tuning which some of the alternative algorithms do not (for they are self tuning). And that does not even address the reality that CMTS firmware releases have a very long gestation period inside the manufactures and the MSOs before being widely deployed. In any case, the AQM mentioned here are uploads at the CPE, and no known (or at least acknowledged) MSO has it enabled on the CPE. Of course, if the OP could show the actual configuration they are receiving on their orange variant we might know for sure. |
|
2 recommendations |
said by gtb:said by telcodad:So, if AQM is a D3.1 "requirement" how did those CMTSs/plugins end up receiving D3.1 certification from CableLabs? CableLabs certifies CPE. ... OK. However, it appears that CableLabs did hold a series of multivendor D3.1 interops (aka "plugfests") between CMTS and modem manufacturers: Cisco Plugged CMTS Into First DOCSIS 3.1 Interop Was One Of Six Vendors That Joined CableLabs D3.1 PlugfestBy Jeff Baumgartner, Multichannel News - December 16, 2014 » www.multichannel.com/new ··· p/386360 |
|
EnasYorlThieves World join:2001-12-02 West ·Ziply Fiber Motorola MB8600 Netgear RAX15 Asus RT-AC66
1 recommendation |
to telcodad
said by telcodad:Thanks, that SCTE presentation from Viavi is excellent! Yeah Page 17 onward was the bomb for me. |
|
gtb Premium Member join:2016-05-16 united state
3 recommendations |
to telcodad
said by telcodad:OK. However, it appears that CableLabs did hold a series of multivendor D3.1 interops (aka "plugfests") between CMTS and modem manufacturers Absolutely. CableLabs is considered an honest broker to define/compare/contrast/evaluate standards and practices, and bring everyone together, and in addition to plugfests, CableLabs has, in their labs, likely have one of most major vendor devices (and a bunch of dedicated testing equipment) to rent and allow vendors to test in environments they might not have access to. But "certification" has a specific meaning in the process, which was the point I (apparently) made badly. Sorry. |
|
1 recommendation |
to gtb
said by gtb:said by telcodad:So, if AQM is a D3.1 "requirement" how did those CMTSs/plugins end up receiving D3.1 certification from CableLabs? CableLabs certifies CPE. Some vendors of CMTSs have "always" supported various QoS (which results in QoE) configurations, but depending on the specific firmware releases it may not support all the variants that are in the entire AQM portfolio (and it is not required to support all in any case), and some of those long known QoS variants can require a *lot* of careful tuning which some of the alternative algorithms do not (for they are self tuning). And that does not even address the reality that CMTS firmware releases have a very long gestation period inside the manufactures and the MSOs before being widely deployed. In any case, the AQM mentioned here are uploads at the CPE, and no known (or at least acknowledged) MSO has it enabled on the CPE. Of course, if the OP could show the actual configuration they are receiving on their orange variant we might know for sure. On my MB8600, I have a fairly constant low level of packet loss (like 0.01% or so) when uploading, and as a result, never get high latency 'bufferbloat' ... I had assumed that this was as a result of PIE AQM ... Do you not have the same behavior? Is this actually just a sign of a low-grade upstream problem? |
|