Making the big push
As the growing ranks of overbuilders construct and deploy fast, high-capacity networks, joining the competitive threat posed by DBS players fashioning high-speed data, time-shifting and interactive services, the buzz they're generating can make it appear to an outside observer like the cable industry is standing still while it watches its own networks grow gray, and long in the tooth.
Nothing could be further from the truth, however, as the industry is on the verge of making one powerful thrust to make sure it reaches the summit first.
Indeed, competition and customer demand for new services are giving the cable industry all the motivation it needs to quicken the pace with upgrades and the deployment of new services that are in high demand.
The tricky part, of course, is reaching that point of deployment. In the middle there lies elbow grease by the barrel and enough deep, innovative thinking to drive a self-prescribed genius to tears.
Conventional cable network wisdom and the days of the 1,000-home node will soon be reduced to rubble within the next few years. In its dusty stead, cablers will make the shift toward smaller and smaller nodes as fiber plunges deeper and deeper.
Here we'll take a look at the trends and innovations that are driving tomorrow's cable networks—infrastructure that will deliver new services like video-on-demand (VOD), telephony, high-speed data and interactive television (ITV). Those trends include hubless and deep fiber architectures, and a controversial "flat network" concept for HFC that carries its own laundry list of pros and cons to pore over. Finally, we'll peer inside Enron Broadband Services, whose "intelligent network" bypasses a slow, congested public Internet—an arrangement that could enable MSOs to offer even loftier data speeds to cable modem customers.The evolution of HFC
As cable operators prepare their networks for existing and future services, one future-proofing question always gallops to the forefront: how much bandwidth will be enough?
If you listen to Harmonic Inc. Director of Market Development Chris Bonang explain it, the answer is a moving target.
"People will use as much bandwidth as you give them," he says, explaining this point during a session at the 2000 National Cable Show in New Orleans.
Because that answer is nebulous at best, Bonang lists three critical variables to consider when upgrading a legacy network or building a new one: it has to support today's interactive services like high-speed data, telephony and VOD; it must scale for future bandwidth needs, because if a cable system can't support some new, popular service, some other company's will; and it must be rapidly deployable.
To meet those requirements, Bonang says there are two architecture options readily available: Hubless and deep fiber.
In a conventional network, he says, a cable operator might have a large hub with a Sonet or an ATM ring connecting a headend, which might house CMTSs (Cable Modem Termination Systems), HDTs (Host Digital Terminals) and video servers. That arrangement, he argues, can be expensive in terms of location and maintenance, and, if built from scratch, can be slow to deploy.
In a hubless architecture, though, the large hub is replaced with a much smaller building or a cabinet that holds EDFAs (Erbium-Doped Fiber Amplifiers) and DWDMs (Dense Wave Division Multiplexers) and demultiplexers. In turn, the CMTSs, HDTs, servers and other equipment are moved back to the headend site. The result, says Bonang, is a rapidly deployable system that reduces operation and transport expenses.
"However, you've still got narrowcasting capabilities," he adds, "because we can use DWDM to target individual nodes with individual wavelengths. It's much easier to find locales for these small cabinets, and it's easy to deploy them very quickly."
Bonang admits that DWDM doesn't have all of the functionality of Sonet, but points to a simple fix to that problem. He suggests that cable operators use Sonet for high priority, low bandwidth traffic like telephony, and DWDM for high bandwidth traffic, thus reducing transport costs.
With deep fiber architecture, which can be used independently or in cahoots with a hubless, what used to be a node becomes what Bonang calls a "concentrator," an optical splitter that can serve 10 mini nodes. Very similar to AT&T Broadband's Lightwire architecture, deep fiber nodes serve between 75 and 100 homes.
By adding DWDM deeper in the network and gradually adding wavelengths until there's one per mini node, bandwidth to each customer can be expanded dramatically, Bonang says.
The payoff with deep fiber, he adds, is four-fold. It supports current service requirements; scales northward; consumes less power because RF actives are eliminated; and offers higher reliability because fewer components are used in the system.
Of course, downstream bandwidth is just one concern for cable operators. As unexpected, new applications make inroads, operators should prepare to dilate the data flow in the return path.
"A network trend we're seeing is the data traffic is turning out to be much more symmetric than operators initially anticipated," says Patrick Harshman, Harmonic's vice president of marketing.
Initially, an asymmetric system made sense, because the only data swimming upstream included mouse clicks and other minute requests. However, as content aggregation server applications like the audio-centric Napster and its video cousin DivX gobble up return path spectrum (see sidebar, page 56), a more symmetrical system is needed.
"People are actually turning out to be content providers themselves," says Harshman. "Napster is one thing. We think video is the next big wave."
As low-cost digital video cameras and editing tools evolve, that need will only rise. "They're practically encouraging people to serve video files," asserts Harshman. "Like Napster is doing today with audio, video will be coursing over these networks, not just in the downstream."
So, that begs the question, how can cable operators prepare for the return path onslaught?
Making Napster zealots pay the piper for that upstream bandwidth obviously is one possible remedy, as Quality of Service (QoS) enters the cable modem picture. Another is to push nodes and fiber deeper into the network.Charter: A "Wired World" study
Taking Paul Allen's "Wired World" vision one step closer to reality, Charter Communications in April broke ground in St. Louis, Mo. on a plan to build more than 500 new master headend, primary hub and secondary hub structures throughout its 6.2 million subscriber cable empire.
Dubbed NOW (Networking Our World), Charter's $3.5 billion upgrade initiative will outfit its systems for two-way digital cable, its Pipeline-branded high-speed Internet service, VOD and other "advanced" broadband services.
A key part of Charter's national plan is to implement a master headend-to-hub distribution plan and to pare down the number of its headends from about 1,250 to a more wieldy 500. Each new building will take up to six months to complete, depending on the building's size. Under Charter's current plan, each node will serve fewer than 500 customers.
"(The upgrade) will enable us to deliver advanced services over a much larger customer base," explains John Pietri, Charter's senior vice president of engineering. "By the time we're done with our consolidation, we'll have reduced down to about 500 headends. We'll still have primary and secondary hub facilities in a lot of the same locations, but the ease of operation and the amount of equipment we'll have to buy to deliver these services is much less and much easier to operate."
By increasing each system's capacity to between 750 MHz and 860 MHz, Charter's upgrade plan is daunting in scope. The construction of its new facilities will complement the MSO's ongoing upgrade of about 75 percent, or 135,000 miles, of Charter's 180,000-mile network of aerial and buried cable. Charter expects to upgrade about 45,000 miles of plant this year alone.
Originally, Charter's upgrade timeframe was to complete the construction within three to five years. Today, however, the MSO expects to finish on the low side of that estimate.
"As we sat down and formulated the plan, the second phase of it was to evaluate what time frame we could do this in," says Pietri. "Since we're spread over a large geographic area, we're not impacting the same people constantly. We formulated what we could do in the first year, and thought we could actually finish it in a three-year period if we had enough capital to fund it."
There are two sides of that issue, he adds. Though Charter must take a financial hit and mete out the required capital to fund the massive upgrade, there are also other benefits to consider by finishing faster and getting new services launched, namely a crop of new revenue-generating services and a powerful retention tool.
At the same time, Pietri says Charter is taking future-proofing measures to ensure it has some network wiggle room for the foreseeable future, which, in cable terms, equates to at least half a decade.
"We're running six fibers to each node, giving us the ability to subdivide those as demand for services increases and traffic dictates," he says.
As for the buildings, which range in size from between 2,600 and 5,500 square feet, Charter has done some number crunching of its own to estimate the rack space needed in order to serve each system size.
"We are designing the buildings so that they can be expanded if new services come about and more space is needed, but that's longer than a five-year plan for us," says Pietri.
Within those five years, expect Charter to add IP telephony to its service repertoire as it deploys more security, fire protection and powering solutions over its networks.
"We're in the middle of installing a system this year that will be testing Voice-over-IP (VoIP)," Pietri discloses, though where those trials will be still remains a secret. What's not shrouded is that the tests will take place in a group of clustered systems that already has 750 MHz of bandwidth and less than 500 homes per node.
If the IP telephony test is successful, Charter plans to launch the service commercially in a number of markets in 2001.Flat Networks: They're fast, but...
While deep fiber and hubless architectures tend to be the topic of discussion at trade shows, others in the cable industry are exploring other innovative (and sometimes controversial) approaches.
One of those approaches is flat networking. Flat networks, as defined in the Local Area Network (LAN) or Wide Area Network (WAN) domain, uses bridges, hubs and OSI Layer 2 switches rather than more complex router technology, the base of today's more conventional hierarchical configurations.
In a cable environment, flat HFC networks would reduce the electronics at the node. "Typically," explains Ham Mathews, director of marketing for ADC's Cable Access Systems Division, "a cable network would have a number of amplifiers coming off of it and serve about 500 homes for the node. If we reduced that to 50 or 70 homes per node, you wouldn't have the cascade (of) amplifiers."
Moreover, a flat HFC network tries to decompose the function of the router, putting something that looks like a switch out at the node. "Because it's switched, it runs very fast," says Mathews.
Protocol-specific flat networks are also known to be easy and inexpensive to configure, reasons why Bill Bauer, the president and CEO of InterTECH and WindBreak Cable, is taking a close look at them.
The struggle with routing technology, according to Bauer, is the process of transmitting packets through a network. "We're starting to see that when you go through each of these routers it adds between 10 and 50 milliseconds of delay," he says. "When you start adding it all up, your delays are pretty substantial—300, 400 or 500 milliseconds—well beyond the 200-millisecond latency threshold for a service like IP telephony."
Employing a technique called subnet masking, Bauer advocates an architecture that uses the intelligence in the TCP/IP protocol to route the packet and tell it where to go. "It starts acting like a router, but it does it in a different way so that it doesn't add latency, because it's not going out and looking up a routing table," he says
While latency could become an issue for router voice services, vendors today are building faster and faster equipment to combat that problem. Those next-generation routers, explains Mathews, are designed to handle interactive, high-speed IP services. "They're dropping latency way down, and they're putting the routing table right there on the line card."
While conceding that faster routers could bridge the latency divide, the solution will hit operators in the pocket book, argues Bauer. "There's a cost associated with that. It's very expensive."
But, just like anything that sounds too good to be true, there's a catch to consider. A flat network's scalability is questionable, because it tends to suffer from broadcast traffic congestion.
"(Flat networks) are blazingly fast, but you can't put a lot of homes on there or the whole thing starts to fall apart," says Mathews.
Troy Wendt, director of marketing at Cisco Systems' Cable Products and Solutions Group, agrees. "As you add devices, at some point the network becomes so noisy that you can't run the standard communication. Not only does it load down your local links, but what's even more problematic is that a flat network will load up your WAN links, which are expensive and a very limited resource, with unwanted broadcast messages.
"If you're looking at a network you want to scale and turn into a multi-service network, I think going with a flat design is going to be problematic," says Wendt.
So, if the scalability of a flat network is rather dubious, who would try such a thing? Well, RCN, a proverbial thorn in cable's side, is one on the short list.
RCN is using the concept to tap the speed it offers, but its customers will have to spend a lot of money on bundled services to make the model work. Basically, RCN has to hunt down the creme of the crop, and target homes that will spend more on communications services than the average home.
"Below that, and they'll end up with a very costly, high-performance network that doesn't pay for itself," says Mathews.
While RCN tangles with the issues in dense markets, Bauer, who serves small- and mid-sized cable markets, says subnet masking can help alleviate the scalability issues with a flat network, though it requires more network management to execute.
"It's more work," he concedes. "But if it's the difference between doing IP telephony and not doing IP telephony, I'll do the extra work."Sidestepping the Internet
While building faster, more robust networks are top of mind with most MSOs, fiddling with its own infrastructure might not be the only avenue to travel on its way to that goal.
Enter Enron Broadband Services (EBS), an offshoot of electricity and natural gas magnate Enron Corp. Through contracts with more than 25 content providers and 20,000 route miles of fiber optics, EBS claims its "Intelligent Network" can deliver high-bandwidth applications like streaming media by circumventing the clogged and congested public Internet.
"We're a wide area network looking for metropolitan partners and subscribers,"says Bobby Robertson, EBS' director of business development, broadband delivery program.
While it has inked interconnection deals with companies like US West, BellSouth and RCN, EBS also has its eye on cable operators.
"We would provide a Layer 3 IP handoff to an ISP or cable modem company at the edge of their network fabric in order to reach their customers," says Robertson. "By working with us, we're providing broadband content to cable modem subscribers without having an impact on the cable network's backhaul or backbone network."
Essentially, EBS would pull its network to a cable operator's edge PoP and install its own servers, routers and switches, and then cross-connect to the operator's edge customer activation router, with "virtually no" incremental infrastructure cost.
By tapping the Enron Intelligent Network, he adds, cable modem service providers can provide speeds up to 50 times faster than the public Internet, the company says.
Today, EBS' network is fairly ubiquitous in the United States. Cable operators can connect directly to the Enron Intelligent Network in major metropolitan areas, or link to it via satellite in rural areas.Fiber-to-the-home: The next step?
While several network-enhancing techniques are being examined, what's the next hill to climb?
After falling off the face of the earth about three years ago, the concept of fiber-to-the-home is again gaining strength.
"Five years ago when fiber started entering the HFC world, some people were saying fiber-to-the-home would be a logical next step," says Harshman.
Noting that fiber-to-the-home is "imminently" more cost-effective than in days past, Harshman believes it still needs more incubator time.
"If you look at the trend today," he adds, "it would suggest that we would be there, or nearly there, in about three years. We're spending more and more time thinking about it." Perhaps thinking about cable's next big push?