The promise of a converged, all-IP future is one of the big drivers cable operators—particularly the larger ones—are contemplating as they cobble together massive, high-capacity broadband backbone infrastructures.

To Build or to Buy?

While some have had that groundwork completed for a number of years, others are polishing theirs off or just getting started. No matter the case, they are all looking at how these backbones will be used to do much more than shuttle around data and support high-speed Internet services. They are also looking to complement that with support for IP telephony and advanced video services and applications.

In addition to handling those efforts, these backbones are also designed to help with the bottom line, and to help them obtain better cost efficiencies by using components that cover all regions, and save them from having to duplicate them at each region, says Gaylord Hart, the director of MSO business and architecture development for Meriton Networks.

While operators didn't have much need for backbone networks 10 years ago, "now it makes economic sense, plus they want that control," adds Andy Audet, vice president of Juniper Networks' cable products business.

Here we'll take a look at a variety of paths operators can take as they divine backbone plans, weigh their pros and cons, and explain how two of the industry's biggest operators are knitting (or have already knit) theirs together. Further on, industry experts will explain how these new backbones could open up a new era for video services.

Approaches to the backbone

Operators, explains Randy Dunbar, the SVP of long haul services for Level 3 Communications, have three approaches at their fingertips: They can continue to buy IP connections; buy wavelengths that interoperate with peering points; or buy dark fiber and light it with their own equipment.

In the IP transit bucket, an operator is simply buying a connection to the Internet where they can send all of their traffic. Before operators were forced in the 2001 timeframe to branch off on their own and scramble to find partners like Level 3, many of the larger ones obtained this capacity from what was then @Home.

Once capacity hits a certain volume, it then makes sense for operators to work out peering links. Operators can do this on their own, but they tend to have less expertise in this area. Plus, "it's hard to grow and maintain those peering links in the backbone," Dunbar argues. Cable operators "work in local areas; they're not used to operating large backbones."

But as operators build direct peering strategies, buying wavelengths and buying dark fiber start to come into play as well.

With wavelengths, operators can obtain intercity capacity from backbone service providers and, in essence, build their own backbone and peer in some locations. But even that approach is complex and requires a big, expert staff to carry out, Dunbar says.

But some operators, Comcast Corp. among them, are also starting to give more thought to dark fiber and running it between the cities themselves.

Dunbar argues that buying up dark fiber can be risky, partly because it is difficult to predict the capacity requirements for those routes.

"You have to make a large, upfront fixed cost investment before you know what your traffic will be or where it will show up," he says.

Dark fiber is also more expensive than it was in the earlier part of the decade when the backbone carriers were in distress. Plus, the cost of dark fiber is miniscule compared to the cost of lighting it and managing it.

With wavelengths, operators tend to have more flexibility and can likely more easily travel the upgrade path.

Dunbar likens this to PCs in the consumer world. Eventually people have to upgrade memory, add processing or buy an entirely new system to keep pace with new applications and other requirements. In the backbone example, there are 10-Gig wavelengths and newer 10-Gig Ethernet wavelengths. Now, Dunbar says, eyes are being drawn to 40-Gig wavelengths.

"Expertise is required," Audet agrees, "but as MSOs offer more and more services, having control over that, I think, is pretty important to them. Certainly, the top MSOs will control the network themselves."

Comcast has been building out its Converged Regional Area Network (CRAN) for about two and a half years. And the company believes it has the general operational and engineering wherewithal to support the approach.

"We have a lot of experience with dark fiber," says Kevin McElearney, Comcast's vice president of backbone engineering and architecture. Comcast's strategy "also gives us the ability to scale to the volumes that we need."

John Leddy
Comcast, which is using Level 3 as its main provider, has 19,000 miles of dark fiber and 10-GigE connections (with 40 Gig on the roadmap). Today, this backbone covers about 95 percent of Comcast's territories.

"We've been lighting that up," says John Leddy, Comcast's SVP of network engineering, noting that the backbone should be up and running, coast-to-coast, later this summer.

Also on the roadmap is support for IPv6, which will help Comcast supply IP addresses and manage the sheer volume of IP devices (including set-tops) that will be hanging off its network.

Comcast has teed up a full set of converged services for the backbone that include linear video with multicast, video-on-demand transport, and support for digital simulcast.

In addition to general VOD assets such as movies, the platform will also be used to distribute advertising. Moreover, it can also be used to bring in content from local markets and distribute them to other Comcast properties around the country.

Leddy adds that one key driver was to provide those capabilities nationally, and not have to duplicate them on a market-to-market basis.

For Cox Communications, meanwhile, backbone building and buying is relatively old hat. It's had an interconnected backbone in place since 2001, and most of it is now served by 10-Gig links. But rather than buying dark fiber and lighting it themselves, Cox decided to lease wavelengths from companies such as Level 3 and one of Level 3's latest acquisition targets, WilTel Communications Group.

Taking a more buy rather than build approach will save Cox having to purchase a bunch of optical electronics, which could hit the tens of millions of dollars for a company the size of Cox, explains Jay Rolls, Cox's vice president of telephone and data engineering.

Figure 1: Internet Backbone
Figure 1: Cox has built and bought its backbone through a wavelength strategy.

Once a company decides to build the backbone, "now, you're in a different business," Rolls says, noting that the MSO then has to spend more time and effort maintaining the health of the optical network. He notes that Cox will keep looking at the build approach and take it only if it makes financial sense to do so.

Now what?

With backbones in place, operators are seeking ways to leverage them for applications and services that extend well beyond traditional data.

"What is changing is that it's not just data anymore. It's VoIP and video back in the core," Audet says.

To help operators push the needle on VoIP peering—a technique that will allow operators to sidestep the PSTN (public switched telephone network) and avoid the carrier fees that go along with it—CableLabs issued a request for information (RFI) on that very subject late last year, but has been relatively quiet about it ever since.

Cable's R&D house said recently that it received more than 30 responses to the RFI and has reviewed and evaluated the technical information that came in. CableLabs added that it is now "evaluating various architectural and operational models for a peering infrastructure," but did not provide a specific timeframe for that.

Today, Cox has four services running on its backbone: high-speed Internet, voice, business services and a dash of video.

High-speed Internet "just dwarfs everything. Everything else is just a blip compared to the flows of high-speed Internet," Rolls says, adding that overall traffic is doubling every year.

He sees video joining the backbone "at a fairly measured pace." VoIP peering, meanwhile, is a bit further out on the horizon for Cox, since the bulk of its phone customers presently use circuit switched technology. "There's not a high financial incentive at the moment [for VoIP peering], but that will change over time," Rolls says.

Although it's not yet true for all operators that own or operate backbones, video is also expected to drive plenty of that traffic in the future.

"We're moving quickly from high-speed data backbones for aggregation and peering and VoIP to a truly convergent infrastructure that can distribute multicast video between markets," says Paul Bosco, vice president of cable and video initiatives at Cisco Systems Corp. This represents "a major shift of what we think of the video distribution model," he adds. "The architectural foundation at a network level is really changing."

That could also significantly complement how digital video presently is delivered over satellite, and perhaps be used to pick up every channel in the country, and generate channel lineups that number in the thousands.

Over time, it could evolve into a fabric that supports new features such as "place-shifting," which allows consumers to view on-the-road video (including linear networks) that is traditionally tethered to the TVs in their homes.

Perhaps just as important is the notion of multi-format simulcasting for regular TVs or video-capable mobile devices. Via one common IP fabric, operators can "format-shift" and transcode the same video for high-end HD offerings or for smaller screens on cellular phones—depending on the bandwidth available and the codecs supported by the end device.

This same architecture could also be tapped to support a growing video "contribution component" set afire by services such as YouTube, which allow users to post and share their own videos.

The architecture being developed by Comcast, for example, is scaling to support thousands of traditional on-air channels, but also being made to collect and move content that originates from subscribers. By leveraging a fiber backbone for such purposes, operators won't be tied as much to satellites and their more limited transponder capacities.

"You can then repackage a 1- hour, 37-minute event, or anything that is a complete broadcast channel, or even a Webcam offering content," Bosco says.

"It's creating a new paradigm far beyond the linear lineup, and into a video world that is much more personalized," Bosco says. "We're moving quickly from high-speed data backbones for aggregation and peering and VoIP to a truly convergent infrastructure that can distribute multicast video between markets," Bosco says.

Although the backbone is an old story from Cox's perspective, it's still a strategic tool and "a great story," Rolls says. "It's a phenomenal asset and will continue to be [leveraged] as more services go to IP."


More partners on deck for Comcast's OTI

Comcast is fueling its converged national IP backbone with the Open Transport Initiative (OTI), which was introduced late last year, and so far includes support from two major routing and optical networking suppliers: Cisco Systems Corp. and Nortel Networks.

At a high level, the OTI will separate out the IP and DWDM equipment, and allow Comcast to mix and match gear on the network. In this case, thus far, Comcast is matching a Cisco router (CRS-1) with a Nortel transport system that includes the vendor's Common Photonic Layer (CPL) and Optical Multiservice Edge (OME) 6500 system. While those routing and optical elements can also be integrated by the same vendor, when they are separated out there is a chance that the two vendors could implement standards differently, and that can add to the complexity.

The OTI should also solve the potential problems of "alien wavelengths," which occur when wavelengths from different vendors are mixed on the DWDM network. Without the proper "spectral purity," those can interfere with other wavelengths on the network, which need to be adjusted to the same level to be optimally set for power and other levels that need to be consistent.

Different vendors can sometimes require different optical levels to be optimized for signal-to-noise ratios, Meriton's Hart explains. "Alien wavelengths may be working, but it might be difficult to optimize them. It's not a simple problem to solve."

It will also enable Comcast to hook into other transport systems and to use alien waves.

Although Nortel and Cisco are the only announced vendors with a relationship to the OTI, Comcast expects to announce others, Leddy says. Comcast, he adds, is also open to offering some of its ODI work to the larger MSO community.