& cplSiteName &

CableLabs Unleashes Specs for Low-Latency DOCSIS

Jeff Baumgartner
6/20/2019
50%
50%

Looking to optimize applications such as virtual reality and online gaming, CableLabs has quietly released an annex to DOCSIS 3.1 that can support sub-1 millisecond latencies.

According to CableLabs, existing DOCSIS 3.1 equipment will be able to support low-latency DOCSIS (LLD) with just a software upgrade.

"Although VR is still struggling to gain widespread adoption, that low and reliable DOCSIS latency will be a boon to gamers in the short term and will enable split renderings of VR and augmented reality (AR) in the longer term," Steve Glennon, distinguished technologist of the Advanced Technology Group at CableLabs, explained in this blog item posted Thursday about the relatively new capability.

A CableLabs official said LLD was included in an update to D3.1 (for the MAC and upper layer protocols interface portion of the spec) in January and has been updated since, most recently in April.

Low-latency gaming is already being eyed as a potential broadband service enhancement and potential new revenue driver. Cox Communications, for example, is testing a low-latency gaming tier for PCs in Arizona that costs an extra $14.99 per month.

Low-latency DOCSIS is a component of the cable industry's broader "10G" initiative, which is focused on next-gen capabilities, such as symmetrical speeds of up to 10 Gbit/s. Although low-latency DOCSIS focuses on HFC networks, CableLabs has stressed that 10G is aimed at a wide range of access network technologies, including FTTP and wireless. Per a new "timeline" for 10G, the current version of DOCSIS 3.1 is considered version 1, with future versions -- 10Gv2 and 10Gv3 -- poised to add Full Duplex DOCSIS and low-latency capabilities, as well as the use of capacity at frequencies well above 1.2 GHz.

In a deeper dive into the LLD effort, CableLabs notes that people tend to detect delays at about 20 milliseconds. So network latency needs to be cut down to near-zero to enable real-time digital experiences and to support apps where low latencies are critical, such as VR, gaming and even telesurgery.

CableLabs says its angle with low-latency DOCSIS involves the optimization of traffic flows, giving latency-sensitive applications a higher priority without slowing down all other traffic.

CableLabs also claims that LLD can be enabled with a "cost-effective software update" (it's not talking exact costs) without "overhauling" the existing HFC network. The organization notes that DOCSIS 3.1 networks already support an Active Queue Management (AQM) feature that improves latency, but LLD aims to take that capability to a new (and lower) level.

According to Glennon, CableLabs launched its low-latency journey more than four years ago as it started to look at apps that will drive the need for sustained speeds of 60 Mbit/s and higher, leading to an examination of "immersive video content" and emerging holographic-based content.

Related posts:

— Jeff Baumgartner, Senior Editor, Light Reading

(5)  | 
Comment  | 
Print  | 
Newest First  |  Oldest First  |  Threaded View        ADD A COMMENT
brooks7
50%
50%
brooks7,
User Rank: Light Sabre
6/21/2019 | 4:59:37 PM
Re: Opening a hornet's nest
Well, depends on how you look at it.

Cable prioritizes its voice differently than it does with OTT voice.  I could even argue that the use of bandwidth for video takes some prioritized bandwidth away from internet services.  The allocation of spectrum on a cable system is a choice, it is not a law of physics.  I could argue that this is essentially paid prioritization on the access network today.

seven

 

 
Jeff Baumgartner
50%
50%
Jeff Baumgartner,
User Rank: Light Sabre
6/21/2019 | 4:29:31 PM
Re: Opening a hornet's nest
Yes, I imagine it won't be too long before network neutrality advocates chime in on this one, as we also expected when Cox started testing its low-lag gaming service. It was interesting to read "priority" language in the documentation though it is quickly paired with an explanation that it's done without slowing down all other traffic. And if any MSO adopts this at some point they will have to be transparent to a fault about what they are doing or not doing. JB 

 
Duh!
50%
50%
Duh!,
User Rank: Blogger
6/21/2019 | 3:06:32 PM
Opening a hornet's nest
Low Latency DOCSIS makes a lot of sense from an application perspective. It's also canonical "paid prioritization" (or would be if "paid prioritization" had a canonical definition).

Expect howls from Free Press, Public Citizen, Free the Internet et al.
Duh!
50%
50%
Duh!,
User Rank: Blogger
6/21/2019 | 3:06:32 PM
Opening a hornet's nest
Low Latency DOCSIS makes a lot of sense from an application perspective. It's also canonical "paid prioritization" (or would be if "paid prioritization" had a canonical definition).

Expect howls from Free Press, Public Citizen, Free the Internet et al.
Jeff Baumgartner
50%
50%
Jeff Baumgartner,
User Rank: Light Sabre
6/21/2019 | 9:33:12 AM
Deeper look at the low latency DOCSIS details
A lot more technical detail about LLD is available in the specs (contained in this PDF), but they point out that the key objective is to "achieve a 99%-ile round trip time of 1 millisecond for packets traversing the DOCSIS network." Another objective is to do this without impacting the other data travelling the network.

A key addition here is support for a "proactive scheduler" that is explained to be a little less bandwidth efficient but a method that provides extremely low latency service. 

While AQM introduced in D3.1 improved median latency and reduced "buffer bloat," the new LLD technology added aims to improve queuing latency, where possible, depending on the behavior of the traffic.

That all ties into the notion of a "dual-queue" mechanism for all traffic:
- One deep buffer (with AQM) for queue-building traffic: this is known as the "Classic Queue".
- One short buffer (with AQM) for non-queue-building traffic: this is known as the "Low Latency Queue".
- A mechanism for scheduling and balancing congestion across the two queues.
- A mechanism for assigning traffic to the appropriate queue, ensuring that queue-building traffic is being assigned to the Classic Queue.

--JB

 
Featured Video
Flash Poll
Upcoming Live Events
September 17-19, 2019, Dallas, Texas
October 1-2, 2019, New Orleans, Louisiana
October 10, 2019, New York, New York
October 22, 2019, Los Angeles, CA
November 5, 2019, London, England
November 7, 2019, London, UK
December 3, 2019, New York, New York
December 3-5, 2019, Vienna, Austria
March 16-18, 2020, Embassy Suites, Denver, Colorado
May 18-20, 2020, Irving Convention Center, Dallas, TX
All Upcoming Live Events