With the demand for a faster Internet connection growing seemingly every day, the cable industry earlier this year settled on an overarching framework for channel bonding.
The practice will give MSOs the ability to “bond” four 6-MHz channels together in order to offer subscribers 100 Mbps of speed. That will come in handy for the expected showdown with telephone companies, but the underlying work on channel bonding will provide a series of other benefits beyond a competitive edge in the marketplace.
IN WORKS FOR A WHILE
Work on channel bonding actually began several years ago, when Cable Television Laboratories Inc. was putting the finishing touches on Data Over Cable Service Interface Specification 2.0, said Doug Semon, vice president of technologies and standards at the advanced engineering group of Time Warner Cable, during a recent CED magazine Web cast on channel bonding.
“After DOCSIS 2.0, we started looking at next-generation architectures and channel bonding to get greater aggregate bandwidth,” Semon said. “The competition is ratcheting up the burst bit rate. The Internet usage pattern is more, more, more.”
Internet-protocol television and video telephony could start taxing DOCSIS 2.0’s capabilities, he added.
“Bonding is inherently extensible,” said Semon. “It’s a cost-effective way of extending our capabilities.”
While the endgame was to provide greater speed to subscribers, channel bonding carried with it other benefits.
“It makes the [cable-modem termination system] modular and that will simplify [hybrid fiber coaxial] installations,” Semon said. A modular CMTS allows an MSO to separate the MAC (media access controller) and PHY (physical layer device) layers in that piece of equipment. That will reduce headend costs and gain flexibility in terms of sharing channels with Moving Pictures Expert Group transport, he said.
Channel bonding also will allow MSOs to move to IPv6 protocols, Semon said, allowing for expanded IP address space. Cable systems sometimes have a hard time managing the IPv4 address space, he said.
“IPv6 obviates the need for that. It’s the way the Internet will be going forward.”
While channel bonding appears technically straightforward, there are some nuances to making it work in practice, said Gerry White, chief technologist in Motorola Inc.’s Connected Home Solutions division.
“Channel bonding is basically the inverse of multiplexing,” he said. “You take low-bandwidth links and make them look like it is one high-speed link.”
The net result: “You’re sending a DOCSIS frame across multiple RF channels in parallel. You need a standard to do that. The standard will logically bond together. In the real world, it is series of packets and the cable modem puts the packets back together.”
The packets aren’t all necessarily the same length, White said. “Smaller packets will arrive first, and the cable modem won’t know which packets to put first. You need the CMTS to add a sequence header.”
That header is added to each packet, he said. “The cable modem looks at the header, then knows what order to send the packets to the PC.”
The same is true in reverse. “The modem adds the headers to the packets, then sends it upstream,” White said. “The CMTS sees the headers, figures out the order, then sends them out to the link.”
White said that today packets go into a service-flow classifier at the CMTS. Those packets are then put into queues, depending on the quality of service required.
“That’s why voice packets get priority,” White said. “But channel bonding complicates that.”
An added distributor function in the CMTS is required, which will take the bonded packets and hand them off to RF links. “The cable modem will get packets for it and for other modems,” White said. “It needs to select the packets destined for it.”
“It looks easy but you need to set up bonding groups and channels,” he said. QOS will be necessary for both bonded channels and traditional channels. And it will require that multiple channels be tuned and demodulated at the modem, he said.
MODEM PRICES WILL FALL
While cable modems capable of handling channel bonding today are costly — selling for more than $100 — prices are expected to drop due to a number of factors.
One is the ever present Moore’s law, said Tom Quigley, senior director of broadband architecture at Broadcom Corp.
A second factor: silicon tuner integration implementations, which will reduce the cost of multiple tuners inside the modem. “Multichannel tuners for channel bonding are coming and will be integrated into CPE [customer premises equipment] chips,” Quigley said.
The widespread use of Gigabit Ethernet switching also spurs the market.
“Integration and switching gives you the ability to be modular,” Quigley said. “You can make chips more modular.”
Operators could also see gains in statistical multiplexing with bonded channels. A 15-megabit channel can handle 11 “sessions,” such as a video-on-demand order. But a 30-Mbps channel could handle 26 sessions, he said — an 18% improvement — while a 45-Mbps channel could handle 41 sessions, a 24% improvement over three parallel channels.
“It’s a real benefit,” Quigley said. “We can also support more users on a given amount of bandwidth as long as that data is sufficiently bursty.”
SHELF SPACE MANAGEABLE
In order to bond channels together, cable will need to find more 6-Mhz channels, at a time when digital simulcast is gobbling up more bandwidth. But Semon said the problem is solvable. “We’re more than capable of handling our shelf space,” he said. “Over time, we’ve proven that.”
In some cases, if could be an even trade, as advanced codecs will provide cable bandwidth efficiency down the road. Moving some MPEG transport to DOCSIS will free up space, he added. “It gives us the flexibility of using bandwidth that was not usable before in the upstream. When the demand is there, we will find a way to do this,” he concluded.
CableLabs is finishing the final specifications on channel bonding early next year.