Motorola’s proposed DOCSIS IPTV Bypass Architecture, or DIBA, has its share of supporters and critics

One option MSOs have for handling increasing video demand is to offload some of it from MPEG transport to a DOCSIS channel–the IPTV approach. That would mean routing more video through cable modem termination systems, which would compel operators to invest in many more CMTSs to handle the traffic. Depending on how quickly demand develops, that could get very expensive very quickly.

Which begs a question: why not just bypass the CMTS? Cox Communications and reportedly at least one other major U.S. MSO are exploring how to better utilize spectrum and save money by avoiding the maze through the CMTS and delivering on-demand video by cutting a path directly through edge QAMs (eQAMs).

CTV Figure 1

The notion is reasonable on the face of it, but the devil is in the details. How would it be implemented? Is it really cost-effective?

The idea is already pitting CMTS vendors against QAM suppliers. The former argue the scheme is unnecessary; the latter insist it will save money. Of course, QAM vendors will stand to gain with CMTS bypass–they’ll sell more eQAMs. CMTS vendors, meanwhile, might have potential revenues at risk.

The basic concept of CMTS bypass has been bandied about for at least three years. John Chapman, chief architect at Cisco Systems, touched on the concept at the 2005 SCTE Conference on Emerging Technologies (for more on this, see Feb. 2005 CED Issue), in a presentation about opening up the HFC pipe. CMTS bypass was considered for DOCSIS 3.0, but rejected.

But Chapman’s paper inspired Cox Vice President of Technology Jay Rolls to consider the ramifications of CMTS bypass for opening up the HFC plant; he and Chapman subsequently filed for a patent, which remains pending as of this writing.

The subject was renewed at the recent ET 2007 conference when Michael Patrick of Motorola presented (on a panel moderated by Rolls) a paper he co-wrote with Gerald Joyce that states the case for CMTS bypass, and details several possible implementations.

Patrick and Joyce outlined what they’re calling the DOCSIS IPTV Bypass Architecture (DIBA), with the observation that there’s a rapidly growing appetite for on-demand video that threatens to overwhelm cable networks.

CTV Figure 2

Patrick posited a peak-usage Saturday night on a node with 750 homes, 75 percent of which are cable subs (that’s high; he knows), with 50 percent of those being on-demand viewers. He assumes each of those homes is demanding two standard-def streams, or about 10 Mbps of capacity. Multiply through and an operator might require 2.8 Gbps per node, or 73 carriers. Operators typically set aside one carrier for DOCSIS and maybe four for VOD.

Compounding the issue is that rising numbers of on-demand customers will exacerbate the long-tail phenomenon. Operators will have to serve up ever growing numbers of unique streams.

Patrick and Joyce posit the best way to cater to that demand is through DIBA. They propose several ways of routing IP video directly through edge QAMs, all of which rely on creating a tunnel through to the eQAM–a function inherent in the DOCSIS 3.0 architecture.

No matter how CMTS bypass is done, some of the functionality of the emerging modular CMTS (M-CMTS) will have to be transferred to another network system. These functions could be resident in eQAMs, but they could just as easily be conducted at the VOD server.

There is still much work that would have to be done to make DIBA a reality. Nonetheless, Motorola is proposing DIBA be added to the DOCSIS 3.0 specification.

John Mattson
John Mattson, marketing director for Cisco’s cable division, said his company evaluated CMTS bypass when it was first proposed for DOCSIS 3.0, but the company’s modeling did not show much of an advantage. Bypassing the DOCSIS control plane and creating more complexity are key drawbacks of DIBA, in Cisco’s estimation.

Doug Jones, chief architect at BigBand Networks, put it more strongly: “If you look at total cost of ownership, [the] modular CMTS is still cheaper than DIBA.”

BigBand is a big believer in the long tail phenomenon, Jones said, but there’s a more efficient way to handle long tail video. “It’s called switched broadcast.”

QAM vendors Harmonic Inc. and GoBackTV flat out contradict assertions that suggest there’s no cost benefit to CMTS bypass.

For starters, CMTSs tend to cost between $10,000 and $15,000 each, whereas QAMs tend to sell in the $500-$700 range.

Harmonic calculates the cost per bit of sending video through both CMTS and QAM to be 25 times greater than routing it around the CMTS and directly through an eQAM.

“IPTV is narrowcast,” says Gil Katz, Harmonic’s director of cable solutions. “If it’s 25 times more expensive, it’s too expensive.”

Katz agrees that several functions that would have been performed by the M-CMTS would have to be performed elsewhere, but he believes Harmonic’s eQAMs are eminently capable of the task. “For us to do DOCSIS processing does not require additional hardware. DOCSIS encapsulation is relatively simple,” he says.

GoBackTV (Note: GoBackTV purchased the digital video assets of Com21 in the second half of 2003 for an undisclosed sum; Dave Baran, GoBackTV’s president and CEO, is a former Com21 exec.) in February introduced a QAM device it demonstrated as capable of doing some of the work of a CMTS. GoBackTV’s GigaQAM IP performs DOCSIS timing and control information for up to eight attached eQAMs, allowing them to deliver a mix of DOCSIS and MPEG-2 video. The unit also provides DOCSIS timing in order to do CMTS bypass in DOCSIS 1.1 and DOCSIS 2.0 networks.

Ramin Farassat, vice president of product marketing at RGB Networks, said not only could eQAMs perform those functions, but “also multiple router/switch manufacturers have stated that they plan to support this functionality as well as the standard IP routing functions through their devices, so I don’t expect this to be an obstacle in the future in building a fully modular CMTS architecture.”

Farassat ventured to turn the whole discussion on its head. If other systems can do what the CMTS does, he said, the real question “is whether there is a place and need in the market for the modular CMTS approach where the functions of downstream modulation, upstream modulation, DOCSIS MAC generation as well as routing can be handled separately and through physically different devices.”

Whether or not some piece of equipment other than the CMTS can do those functions is not the issue, according to Tom Cloonan, CTO of ARRIS’ Broadband division. “If you’re just moving functions among boxes, it’s not clear to me there’s an advantage,” he says. The same argument was made by Mattson and Jones.

And as for making it part of DOCSIS 3.0, DIBA just isn’t worth the distraction, Jones argues. That ratification process, plus the associated testing and integration work could create a delay in the 3.0 generation of technology by as much as a year, Jones warns.

DIBA also diverges from the technology roadmap. Mattson says Ethernet and IP are “the two strongest trends in the last 20 years. Why is Ethernet winning? It’s dirt cheap and dirt simple. Why is IP winning? It’s completely flexible. We see the modular CMTS as embracing those two technologies, and that’s why we think it’s so powerful.” DIBA is a diversion from that, he adds.

“It’ll drive the network away from convergence,” agrees Cloonan.

Arguments about cost and viability aside, Mattson, Cloonan and Jones all suggested the question of DIBA is academic, anyway. All three said they hadn’t seen much, if any, interest from their customers.

Rei Brockett, VP of marketing at GoBackTV, says the CMTS bypass concept is being embraced more readily by European operators who are under much greater competitive pressure from IPTV providers. Katz of Harmonic says two U.S. MSOs he would not identify (other than to say they are “very big ones”) are evaluating CMTS bypass.

Cox’s Rolls agrees there’s little reason to overburden the CMTS. There are distinct advantages to turning the HFC network into an enormous pipe. Although it would take an enormous amount of work, it would simplify many things, he argues, and CMTS bypass would be part of doing that.

The challenge, he says, is that the system as Cox conceives of it would require customer premise equipment with a receiver capable of receiving 24 QAMs simultaneously. That technology does not exist, though he said at least one silicon vendor, BroadLogic Network Technologies, is close to achieving it.

Cox Senior DOCSIS Engineer Ben Bekele says the MSO has only begun to explore the possibilities in earnest, but it’s already made a convert. Bekele notes he’s had recent discussions about the technology with Broadcom Corp., “who had some of the same reservations as the CMTS vendors, and they walked away ready to help with a spec.”

Cox was talking with Broadcom, Bekele explains, because at the time DOCSIS 3.0 was first being envisioned, Broadcom had come up with some concepts relevant to CMTS bypass.

The specification would have to include what Cox is calling a striping engine–a piece of equipment that could manage video in a CMTS bypass architecture. That functionality could be placed in an eQAM, “which is why they’re happy,” Bekele says. “CMTS vendors might see bypass as a threat, but a CMTS blade could just as easily do it.”

Motorola is pushing DIBA as part of DOCSIS 3.0, which might be as problematical as some CMTS vendors fear, but Cox wants nothing to do with that. “No way,” Rolls said. “3.0 is baked. This is not part of that.”

Bekele says the CMTS bypass spec Cox develops will be introduced to CableLabs “after DOCSIS 3.0 is all settled.”