Docsis Flirts With 5Gbit/s

While Docsis vendors work on refreshing the technology to reach 10Gbit/s downstream speeds, a recent field trial shows that current gear can get almost halfway there.

Ahead of next week's ANGA Cable Show in Cologne, Kabel Deutschland GmbH and Arris Group Inc. (Nasdaq: ARRS) demonstrated downstream speeds of 4.7Gbit/s on live Docsis 3.0 plant connected to a school in Schwerin, Germany.

It's considered a new Docsis land speed record, achieved by bonding 96 8MHz-wide EuroDocsis channels (that's 768MHz total spectrum). The trial used 12 Arris cable modems, each capable of bonding eight downstream channels (the latest generation of Docsis 3.0 modems can bond up to 24). (See Intel's New Docsis 3.0 Chip Guns for 1-Gig .)

To complete the kluge setup, Arris and the operator inserted a Layer 2/3 switch and aggregated and multiplexed those 96 channels into a 10Gbit/s feed linked to a PC. Out on the network, the trial was fed by Arris's flagship CMTS, the C4, which was outfitted with four 24-downstream port line cards (plus one upstream card with 12 ports for good measure).

Some good timing also played a part. KD and Arris conducted the trial just as the operator was about to open up a new fiber node, so there was no other traffic running on that portion of the recently upgraded 862MHz plant. "For a short window of time, they (KD) saw an opportunity to try something," says Arris CTO Tom Cloonan.

The 4.7Gbit/s demo isn't anything near practical, but it does show what the current equipment generation can do. "We think this is just the starting point," Cloonan says, noting that cable modems capable of bonding 32 or 48 channels could be the next possible step, and maybe 64 after that.

The road to Docsis 4.0?
Converting most or all of cable's spectrum into one giant Docsis IP pipe is the sort of thing that Arris, Cisco Systems Inc. (Nasdaq: CSCO) and Motorola Mobility LLC are proposing with a next-generation Docsis that they claim could support 10Gibt/s downstream and 2Gbit/s upstream. But with the number of tweaks and additions on the table, such as spectrum upgrades to possibly 1.7GHz, it amounts to a quantum leap for the technology. (See Does Docsis Have a 10-Gig Future? )

Other ideas aim for nitty-gritty efficiency boosts, squeezing out more bits per Hertz. For example, the Docsis vendors are keen on OFDM (orthogonal frequency-division multiplexing), a modulation scheme already popular in mobile world, that would essentially take today's 6MHz channels (for North American Docsis) or 8MHz channels (for EuroDocsis) and chop them up into much narrower, 10KHz-wide channels. The thinking is that the smaller widths would let cable operators utilize noisy pockets that are otherwise unusable.

Another idea being pushed is low density parity-check (LDPC), a new forward-error correction flavor that would aid data transmissions in noisy plant conditions by reducing the amount of bandwidth overhead required for the current technique, called Reed-Solomon. The combo of LDPC and OFDM could improve spectral efficiency by 18 percent to 24 percent, some believe.

Docsis vs. EPoC
Those techniques are also being considered for EPON Protocol Over Coax (EPoC), an emerging Institute of Electrical and Electronics Engineers Inc. (IEEE) standard that some view as a competitor to a next-generation Docsis platform. (See Huawei Looks Beyond Docsis.)

"There's not a reason to pick one over the other in terms of performance; they're virtually identical," Cloonan says. "But cable will need to look at it from other attributes, other than performance."

Cloonan says backwards compatibility is one area where a next-gen Docsis platform would have an edge over EPoC. Of course, Arris, like Cisco and Motorola, has other good reasons to favor Docsis -- they own the bulk of the CMTS market and have some sizable turf to protect.

But Cloonan doesn't completely discount EPoC as a long-term option. EPoC, he said, could serve as a "stepping stone" for cable's eventual move to fiber-to-the-home, "if MSOs want to prepare for a 2030 timeframe."

Still, Docsis might also be made to play a role in that FTTP transition. In one possible scenario, a new version of Docsis could be paired with Radio Frequency Over Glass (RFoG), a cable industry standard that lets operators pull fiber to the premises while preserving their headends, backoffice and provisioning systems, and cable modems and set-tops.

Cloonan believes most cable operators will go with a next-gen Docsis system "if it's defined." That's still an open question, as CableLabs has yet to make any decisions about what comes after Docsis 3.0. (See The Docsis Addendum and Costs Could Keep RFoG a Niche Player .)

And those decisions may depend on geography. Cable operators in China and India are already using Ethernet-over-coax technologies to serve high-density areas, so a standard like EPoC could cater to those markets, says Infonetics Research Inc. analyst Jeff Heynen.

Heynen does agree that EPoC makes sense as an FTTP transition technology for cable but still thinks Docsis will reign for the foreseeable future. "We're at least ten years away from Docsis exhaust," he says. "And maybe that's conservative."

— Jeff Baumgartner, Site Editor, Light Reading Cable

Page 1 / 2   >   >>
VictorRBlake 12/5/2012 | 5:30:44 PM
re: Docsis Flirts With 5Gbit/s

Actually my mistake. It's 2E24-1 (16M), not 2E8 LLID.

I won't dispute the performance hit to large scale polling, I think the likely uses are small numbers of business customers that will work very well for the EPON MAC exactly as it is today. But even if we were to ponder residential, think about it. Typical DOCSIS networks today, as a reslt of node-splits and smaller serving group sizes are in the same range (tens to hundreds of statations per SG).

As to performance, I think it's hard to argue that somehow the DOCSIS schedulers perform better than the EPON MAC scheudlers. It is literally no comparison (thus the reason EPON is well suited for business SLAs). The EPON scheduler is deterministic (not probabilitistic) and in practice the latencey (and variation in latency -- aka jitter) is FAR lower on EPON networks than in DOCSIS. There are a number of studies both by vendors and which I've conducted in the field (myself) that prove out the latency and jitter measurements on both DOCSIS and EPON.

So despite the truth of the theoretical problems of polling, in practice it (a) is not a problem for the scale of typical deployments and (b) in practice performs better than the alternative in terms of performance.

Duh! 12/5/2012 | 5:30:45 PM
re: Docsis Flirts With 5Gbit/s


Understand your point, but I think you're completely missing mine.

Yes, I understand that you can logically support 255 LLIDs per PON.  And I was, admittedly, working on the basis of the PX10 and PX20 PMD specifications...  forgot about PX30.  My bad.  And I had understood that a media converter could support multiple coax network units - although I would expect it would take the place of a multi-tap which typically supports 2-8 subscribers in a suburban deployment.   But that's beside the point.

What I was getting at was MPCP is a polling protocol:  REPORT MPCPDUs get sent only in response to GATE MPCPDUs, which are unicast.  This is for good reason in a PON;  photodetectors can be damaged by overload.  For best-effort service,  DOCSIS is a contention-reservation protocol: the CMTS issues contention opportunities by broadcasting REQUEST or REQUEST-DATA IEs in the MAP, and CMs contend to transmit Request frames, with backoff in event of collision.  

It's well understood that the performance of polling protocols is more sensitive to number of stations than contention-reservation protocols.  Significant portions of the upstream are consumed with unproductive polling.   Contention-reservation protocols are more sensitive to high traffic intensity, as collisions become more frequent and backoff windows increase.    Mechanisms like piggybacking, and the unsolicited grant and polling services help with this.

So my point was that as MSOs roadmap the evolution of the  HFC/DOCSIS/MPEG2-TS network,  I don't understand how EPON MPCP has any technical advantage over the DOCSIS MAC.  This is not an argument against DPOE in greenfields and enterprise applications.  Just not understanding the engineering case for EPOC.

VictorRBlake 12/5/2012 | 5:30:46 PM
re: Docsis Flirts With 5Gbit/s


If it is possible and feasible to run on the same HFC, although I think in practice that is very complicated because the electrical power budgets are quite different and none of us has studied potential interferrence. It probably won't be necessary, but could be done.

If we talk about a more distant future (after the standard is done and chips have been out for a few years), the market for EPoC in Asia is huge. Just to give you some idea -- the number of EPoC chips at stake (PHY chips) is larger than the total # of DOCSIS 1.1, 2.0, and 3.0 chips ever manufactured in the world and which will be manufactured up to that point (say 2014). So looking at the economics, what *could* happen is that the cost of an EPoC PHY chip + existing EPON MAC chip could be lower than an equivalent bandwidth DOCSIS chip. That is, with a higher volume than DOCSIS chips, the EPoC chips (plural or if combined possibly into a single chip) could be economically compelling. Although the economics compells us to ponder this, it isn't a reality and until the chips are out and the costs known. There is no competition because "EPoC" chips don't exist today (in a multi-vendor interoperable standards way) AND which are cost effective.

Just to be clear, other than network management (visibility of the phy conversion and OAM of it) there is literally nothing to be done on existing OLTs (or DPoE Systems) to make EPoC work. (That's the whole point). This means existing OLT and ONU MAC chipsets are the same -- that is the SAME volume and pricing in the market today. It's a big market. It's not my place to share some of the data I know about, but the global market for EPON and EPoC is >> (much much) g.t. (greater than) DOCSIS, so it stands to reason that the economic benefits of manufacturing scale will matter just as they always have for other IEEE Ethernet technologies.  IMO, Ethernet technologies always win out because of interoperability , volume, and I would argue -- simplicity. Betting against Ethernet doesn't usually work out (read CLM).

Until that time comes, EPoC is (a) in North America -- a value add for EPON, (b) for DPoE in particular, a value add for MSOs (again for biz services), and (c) globally a standard for FTTC/FTTB/FTTP wherein the premise/building (or sometimes the cabinet at the 'curb') is EPON and the remainder is EPoC.

If North American MSO uses EPoC to grow their business service markets, reducing the capex (fiber) construction along with DPoE, they can easily offer many services in the coming years that will pay for the network equipment (DPoE Systems) with a fast ROI because the cost of DPoE (EPON) is not in the gear, it's in the fiber construciton.

Business customers are literally desperate for something beyond a T1 and the monpolistic rates they have paid for those services for years. Compared to twisted pair (VSDL, etc.), EPoC is literally 1-2 orders of magnitude more capable and could enable MSOs to grow their already fast growing business services at a faster pace and lower cost. We are talking about 20x to 50x the current MSO gross business revenue and rightfully taking their 1/3 (in markets with ILEC + CLEC) to 1/2 (ILEC only, no CLEC markets) of the competitive business services market out there. Whether it's e-rate projects, Metro Ethernet services, or Metro-E or IP for Internet or cloud acces, this is a huge business for MSOs and having an option on the table for those multi-hundred Mbps to Gbps customers (which is the EPoC sweet-spot) is very useful.

Finally, there are some vertical markets like CTBH, WiFi backhaul, microcells, etc. that don't necessarily justify fiber construction but require the lower latency and higher bandwidth -- and don't make sense to share on a DOCSIS network. (i.e. 500Mbps cell tower, etc.).

Jeff Baumgartner 12/5/2012 | 5:30:46 PM
re: Docsis Flirts With 5Gbit/s

Hi Victor,
Thanks for the thoughtful analysis on this. As potential usage scenarios go, then, I take it that EPoC, particularly in N. America, is viewed as primarily an option for business services, much the way cable's using EPON today? 

I don't think anyone sees EPoC as a potential replacement for DOCSIS anytime soon (in fact , i have noticed that the potential spectrum plans for EPoC would live alongside what's set aside for DOCSIS and other cable services), though I do get the sense that tension among the two groups is building. But down the road do you think cable will have to make a hard decision even when it comes to residential services regarding whether to stick with DOCSIS or start to move everything to EPoC?

I realize that we're talking future, crystal -ball scenarios here, but all this what-comes-after D3 talk does seem to try to take into account what's on the long-term horizon here. JB



VictorRBlake 12/5/2012 | 5:30:48 PM
re: Docsis Flirts With 5Gbit/s

Two things:


1. DOCSIS is an acronym and as such it should be written in all caps as "DOCSIS", not as "Docsis"

2. The rational arguments for using the EPON MAC for EPoC have nothing to do with DOCSIS. The reason to use the EPON MAC for EPoC is because the purpose of EPoC is to extend the EPON network reach to locations where there is no fiber and where it might not necessarily be needed. Example, you have a building with FTTP/FTTB. But there are no fiber (only coax) risers. Perfect situation for EPoC. Ex: 2. You have an FTTx network for DPoE/EPON for business services. Customers wants service, but fiber construction might taken X months to permit and build, use EPoC -- again perfect solution.

While i understand that AFTER EPoC (IEEE standard) is completed it could one day compete against other technologies, the goal of the project is not to replace anything, but to give operators and vendors a too to extend EPON networks to places where they could not reach previously. As for EPoC versus DOCSIS, the intended products and services are NOT the same (at least not at this time).  EPoC for example, might be well suited for a 1Gbps customer, whereas that isn't a service goal for an individual customer on DOCSIS. Likewise, DOCSIS might be utilized for a service like 30/5, whereas that isn't really the target goal for EPON/EPoC use.

Aside from data rates, DPoE (which uses EPON) includes a comprehensive feature set for Metro Ethernet services which again -- is not the same as DOCSIS. DPoE can offer IP services, but the primary interest for DPoE and EPoC is Ethernet (Metro Ethernet) services.

Constaltly comparing DOCSIS and DPoE/EPON is like comparing CDMA and WiFi or LTE and WiFi. Like wireless (where WiFi and cellular are now understood to be complimentary), so too do operators see them as complimentary, offering different capabilities for different services.


VictorRBlake 12/5/2012 | 5:30:48 PM
re: Docsis Flirts With 5Gbit/s

I also wanted to comment on the splits. I'm not sure where the previous positing's author gets the idea that "16' is the number of devices supported for EPON/EPoC. in fact the field allows 2E8 minus 1. That aside, the limiting factor in EPON (or any PON) is the optical power loss compared to the receiver sensitivity (optical power budget). Current deploymens around the world typically use 32 to 64 (rarely is 16 unless the reach is extradinarly far like 30km+). The ExEPON work in IEEE (active right now) will expand that to much higher optical splits or reach or combination of the two.

That aside, the optical limitaiton only applies to ONUs (we don't call them ONTs in EPON typically). A media converter would phy translate the optical to electrical, so the reach on the EPoC portion is related to electrical power levels. For example, you could have a 1:32 or 1:64 (or a combination like 1:2, staged with 1:16 or 1:32 to get to 32 or 64) and have EACH of those be a media converter. If each media-converter  in-turn supports say 256 coax network units (CNU) that would be 256x32 or 256 x 64 or 8192 or 16,384 cnu's per PON. Now that isn't likely for a variety of power reasons on the coax side, but more importantly it isn't likely because out of a 10Gbps PON, if you are selling services in the 100Mbps + range, you will run out of bandwidth before you come anywhere close to the physical optical or in the case the logical identifiers (LLIDs) for each logical ONU (CNU) even when there are multiple LLID's per ONU (CNU).

Long story short, EPON supports 2E8-1, the optics supports easily 1:32, 1:64 and more, and coax will support a reasonably large number (TBD since the standard is obviously not done yet) which is >> (much greater than) the practical need. Hudreds of CNUs per PON would be sufficient for a 10Gbps EPON networks.



Duh! 12/5/2012 | 5:30:53 PM
re: Docsis Flirts With 5Gbit/s

I haven't yet heard a cogent technical argument as to why EPON MPCP is more optimal for cable evolution scenarios than the DOCSIS MAC.   If anything, the fact that contention is not feasible in a PON, and that MPCP is optimized for something like 16 ONTs per PON segment, would lead to the opposite conclusion.  What I've heard seems to be more in the way of speculation about effects on the ASSP and equipment markets. 

It would be very interesting to do comparative simulation studies of MPCP and DOCSIS MAC, with identical PHYs, in networks with splits ranging from 32 to 250.

Jeff Baumgartner 12/5/2012 | 5:30:54 PM
re: Docsis Flirts With 5Gbit/s <div>

As "it" goes, I'm referring to&nbsp;what's underway for EPoC, versus the&nbsp;proprietary &nbsp;EPON over Coax access technologies that the likes of Entropic already have now. But I do agree with Seven that alot of factors will determine where a tech like EPoC will make sense, and where it will have a hard time taking hold, at least for a long time. JB



AESerm 12/5/2012 | 5:30:55 PM
re: Docsis Flirts With 5Gbit/s

Don't think it's flamed out in Asia, although best to define 'it.' If we're talking EPON protocol over Coax, that's one thing. A more generic, non-standardized Ethernet over Coax, which was the buzz of the ICTC in Xi'an (China) last October according to some who attended, is another thing.&nbsp;

paolo.franzoi 12/5/2012 | 5:30:55 PM
re: Docsis Flirts With 5Gbit/s


There are a lot of "depends" on these things. &nbsp;Access is a funny business where demographics, the age of the physical plant, geography and all kinds of things play into the costs to do various upgrades.

In the end it almost always boils down to an 80/20 decision. &nbsp;Can I do the upgrade relatively inexpensively in 80% of my locations and provide a viable service for a significant period of time?

If I take those questions to the telco world and use U-verse as an example....the big challenge is 2,500 foot loop lengths. &nbsp;This works fine in urban and suburban areas. &nbsp;But the costs are too high in rural areas (in fact at some density FTTH is actually cheaper to do). &nbsp;So, you see AT&amp;T's reaction. &nbsp;There is a significant amount of "non U-verse" properties inside of AT&amp;T.

If I go to cable, they will look at these upgrades on a property by property basis and be driven by the presence of FiOS or U-verse. &nbsp;Interesting to see what the reaction will be to smaller properties is. &nbsp;Verizon has been selling them. &nbsp;AT&amp;T has been ignoring them. &nbsp;Wonder what cable will do?



Page 1 / 2   >   >>
Sign In