Doubling Down on DOCSIS 3.1
DOCSIS 3.1 and PNM were hot topics at Cable Tec Expo in New Orleans and in most major publications and articles for 2015. Comcast is the largest driver of DOCSIS 3.1, but my recommendations are that whether you are a tier 1, tier 2 or even tier 3 cable operator DOCSIS 3.1 should be on your roadmap. Consumer behavior and technical innovations are leading factors behind this statement and I elaborate below.
Consumer Behavior and the State of Internet Traffic Growth
In May 2015, Cisco released the complete VNI Global IP Traffic Forecast, 2014 – 2019[1], which tells us that subscriber traffic across our networks is going to continue to increase – a lot.
As an industry, we extrapolate from this study that our downstream data rates continue to increase on average of 50% each year (CAGR). Depending upon demographics we see this number to be considerably higher in rural areas where Netflix usage is much higher, likely due to fewer nightlife activities (I grew up in a rural setting – so I can attest to this).
Something the Cisco VNI study and our models are not able to take into account are upcoming changes in user behavior due to technology changes. For example, most of the over the top (OTT) set top boxes, such as Apple TV, Roku and Amazon Fire TV now support streaming 4k TV, with 4k content available from Netflix, Amazon Instant Video, and the newer UltraFlix service. If you have a 4k / UltraHD TV or are thinking about getting one, this becomes the only reasonable way that you can get 4k content for your shiny new TV. This is because every set top box deployed today is limited to 1080p or less. Comcast announced in August that they are ready to “flip on its 4k switch” as soon as consumers begin demanding it [4k service] en masse.[2] Its good that there is an MSO response, but will this be enough?
The reality is OTT content is currently the most significant demand on our DOCSIS networks. Netflix HD quality, their existing service which is equivalent to 1080p, consumes roughly 5 Mbps or 3 Giga Bytes (GB) per hour. When users begin streaming 4k videos over Netflix (and other similar services) the increase on DOCSIS networks will go from 5 Mbps to 25 Mbps. That is a 5x increase. The hourly usage increases form 3 GB (HD) to 7 GB for the 4k service. Are you required to have a 4k TV to stream 4k? Nope! As I’m writing this article I’m streaming a 4k video to my Retina Mac Book pro at an average of 20 Mbps with bursts to over 40 Mbps. The upstream is also seeing large TCP/IP acknowledgement activity. Both can be seen in the data tracker shown in the figure from my laptop during a screen capture during a time of 40.7 Mbps download and 20.8 Mbps upload while the video was playing.
Figure: Download and Upload of Netflix 4k Video
Will this have an impact on our DOCSIS networks? Yes – an expected large impact. When should we see the impact hit? This holiday season. A common statement I make to clients (cable operators) is to prepare for capacity increases before the holiday season. Tis the time for new technology gifts. I anticipate this holiday season will be full of new Apple TVs and 4k TVs with many users intent on watching 4k content, which was previously not available. Now it is available and now it will be coming to your DOCSIS network with five times increased downstream usage.
How do we manage this increase in OTT usage?
Why do I say double down on DOCSIS 3.1? Because DOCSIS 3.1 is very efficient with our bandwidth. Let’s say today you have eight (8) DOCSIS downstream channels, which gives you a net capacity of 8 * 37 Mbps = 296 Mbps. You realize that you need to double every two years due to the 50% CAGR and now you are really concerned about this 4k OTT streaming issue and decide you perhaps you may need to plan the next four years a little more aggressively. Using the standard 50% CAGR you will need to increase your eight downstream to 32 downstream (you doubled your capacity over four years). But being an aggressive designer you realize that 4k streaming video or some other technology will cause you to run short on bandwidth, so 32 downstreams in four years will not be enough. What do you do? You are already consuming 192 MHz of RF bandwidth with 32 DOCSIS channels and your capacity is only 32 * 37 Mbps = 1.184 Gbps. Further, to achieve 32 downstream DOCSIS channels you must decide which revenue generating video QAM channels you will eliminate.
Move over DOCSIS 3.0 and move in DOCSIS 3.1. Let’s re-evalutate this scenario if we had the option of a DOCSIS 3.1 downstream in our plant in a two year time frame. Our first migration will be to 16 downstreams resulting in a net bandwidth of 16 * 37 Mbps = 592 Mbps. This will cover us for the next one to two years. But legacy single carrier QAM (SC-QAM) DOCSIS channels have a spectral efficiency of approximately 6.2 bits/sec/Hz at 256-QAM. This is simple math because the net subscriber throughput is 37 Mbps and the channel occupies 6 MHz. Therefore, we obtain the channel efficiency by dividing 37 Mbps / 6 MHz = 6.2 bits/sec/MHz. Now spectral efficiency should not be a confusing term to anyone. DOCSIS 3.1 running at 4096-QAM has a spectral efficiency of 10.5, nearly double that of SC-QAM at 256-QAM. But to achieve this you need an MER at the modem of 41 dB. Many people will feel this is difficult to achieve. However most systems today are running 256-QAM in the downstream, which means the system will support 1024-QAM in DOCSIS 3.1 due to the LDPC error correction. 1024-QAM has a spectral efficiency of 8.6. Using a spectral efficiency of only 8.6 it can be seen that by adding a DOCSIS 3.1 OFDM channel that is equivalent to 16 more DOCSIS SC-QAM channels we will achieve a net throughput of 825.6 Mbps (1 Gbps if we can achieve 4069-QAM). That is a total bandwidth of 825.6 + 592 = 1.4 Gbps in the same bandwidth of the DOCSIS 3.0 only case where we just used SC-QAM DOCSIS channels and only achieved 1.184 Gbps.
Let’s summarize. With the DOCSIS 3.1 CMTS we can still allow all of our legacy DOCSIS 2.0 and 3.0 modems exists with 592 Mbps aggregate bandwidth. As new users and power users come online we can migrate them to DOCSIS 3.1 modems which have between a 825.6 – 1000 Mbps aggregate bandwidth. Overtime we can slowly convert the legacy SC-QAM channels to DOCSIS 3.1 OFDM channels that have higher spectral efficiency. The result of 32 channels converted completely to OFDM is 32 x 6 MHz/ch x 10.5 bits/Hz = 2 Gbps. Remember our original model of 32 SC-QAM DOCSIS channels would only yield 1.184 Gbps. Don’t need 2 Gbps? No problem. Just use less DOCSIS 3.1 bandwidth. DOCSIS 3.1 is quite flexible and allows you to use a channel bandwidth as small as 24 MHz, which is equivalent to four (4) DOCSIS SC-QAM channels, but you get up to 252 Mbps using DOCSIS 3.1 instead of the 148 Mbps you would achieve with DOCSIS 3.0.
Spectral Efficiency is Just the Beginning
If spectral efficiency does has not convinced you by giving you flexibility in nearly doubling your future bandwidth, thus saving you CAPEX and OPEX on future upgrades, then perhaps the built-in-test features of DOCSIS 3.1 just may.
DOCSIS 3.1 was the first DOCSIS specification to include a wealth of test functionality to become standardized in the specification (see section 9 of the DOCSIS PHY specification). Proactive Network Maintenance (PNM) has become a very hot topic and application across the industry. A challenge of PNM pre-DOCSIS 3.1 has been incompatibility with equipment vendors. In the DOCSIS 3.1 specification PNM has become a “MUST” requirement. Further, many new and more powerful features for PNM functionality were added to the specification that we were unable to previously take advantage of, such as the ability to perform a return path sweep using a cable modem and the CMTS. A task, such as sweeping the return, with a DOCSIS 3.1 modem and CMTS can be completely automated with a PNM server and requires zero truck roles.
All DOCSIS 3.1 modems are required to support full band capture, which means they act like a spectrum analyzer in the subscriber’s home. All DOCSIS 3.1 CMTS are required to function as upstream spectrum analyzers, which means you have complete visibility into each return path – and there are many bells and whistles on the new upstream spectrum analyzer, such as triggering and impulse noise detection. However we have noted one CMTS vendor not supporting some critical upstream spectrum analysis features, so be sure to check with your vendor before making your purchasing decision.
Finally, when we look at the fundamental technologies employed in OFDM and OFDMA pre- and post equalization we know that our ability to identify and locate impairments will become 40 times more accurate than pre-DOCSIS 3.1 devices. This is because instead of channels being 6 MHz wide they are created with 25 kHz or 50 kHz wide sub-carriers, each with pre-eq in the upstream or post-eq in the downstream. Post-eq in the downstream means that we can also identify downstream impairment type and location as we currently do in the upstream.
Summary
DOCSIS 3.1 may not yet be on your roadmap or may be further out. But due to subscriber behavior, technology innovations such as OTT streaming 4k TV and new apps we have yet to see, we recognize that there is never enough DOCSIS capacity. DOCSIS 3.1 will provide nearly twice the spectral efficiency of current DOCSIS specifications along with backwards compatibility. This will allow substantial flexibility over the coming years as subscriber traffic rapidly grows and will be sure to protect your next CMTS investment. Further, the built-in test included in DOCSIS 3.1 and PNM is an invaluable tool that will further yield cost savings to your bottom line as well as improved subscriber quality of experience.
As seen in Broadband Library
By Brady Volpe
[1]http://www.cisco.com/c/en/us/solutions/service-provider/visual-networking-index-vni
[2] http://bgr.com/2015/08/24/comcast-4k-technical-network-distribution/
Brady,
Thanks again for a great post on D3.1. However, and I may be in a camp of one on this 🙂 I’m not convinced the BUSINESS case for PNM stacks up. i have been following it for a while and whilst I agree from a pure Engineering point it is the right thing to do (with infinite resources, time and money), I’m not convinced that this PHY/MAC approach is the right one. The theory goes if you fix a fault identified by PNM, that this translates into a fault as seen by the customer at layer 3 and above-so FEC errors/poor SNR/pre-eq working it socks off means poor browsing, high latency, packet loss etc etc. I dont see this, and haven’t seen anything yet that proves this. Once you cream off the chronic customers from PNM (say the top 1%), the point isn’t proved that the other 99% suffer as badly-I would also argue that those 1% could be indentified by the usual DOCSIS stuff-so flap lists, poor DOCSIS pings, high insertion rate, UnCorr FEC count etc.
We really need to start looking at toolsets that identify problems for customers at layer 3 and above-which ties in with the comments you made about Netflix and OTT generally-for me, this should be the focus-not layer 1 and 2 that goes back 20 years, and doesn’t really focus in on the true customer experience of today.
Hi Cabs,
It’s called PNM –Proactive Network Maintenance because it is being proactive and not reactive, Pre-eq is “masking” issues on the plant keeping the end subscriber happy. But, it’s up to the cable company to be proactive to track how hard this pre-eq is working to fix issues that could potentially get worse and then affect the customer later. The cable plant does not heal itself.
There are so many times in history where people were skeptical of a new technology and now a days we cannot image living without it or going back to how it used to be. Thanks for the feedback!
Thanks Mia-with 20+ years in the industry I get the concept of eq and the idea of being proactive :-). My point was that phy issues dont always lead to customer issues.My opinion is that a lot of people in the industry are still using a mindset from 20 years ago, where certain people exalted the use of SNR and Codewords as #bad, whatever the rate,size or occurrence. PNM is just another link in that chain. I do wish that we could get away from the whole layer 1 & 2 as the start and finish of any network maitenance , and start building tools (even a simple ICMP health “heartbeet” from the CM back to the CMTS?) that gives us a better idea of what is happening, in relation to the devices that are connected to/pass their traffic through, the CM.
Ta
Cabs
There are a lot of elements to a cable plant and your point is well taken. Thanks for the comments!