Everyone knows about AT&T Corp.'s grand cable TV plans, byt few realize just how ambitious those plans are. Try this: The infrastructure AT&T is building will ultimately support 64M bit/sec downstream links to the home.
The company won't get there overnight, but the idea is to rewire America with broadband in preparation for the day when communications means truly integrated voice, streaming video and a boatload of data. It is a dramatic about-face for a company that was casting about in the early 1990s looking for its role in a digital future.
"We realized we were a kilobit company in a megabit age," says Norm Schryer, division manager of broadband services research at AT&T's research arm, AT&T Labs in Florham Park, N.J. This 1997 awakening led to a $110 billion cable TV buying spree, giving the company access to a potential of 27 million residential customer homes.
AT&T has already upgraded 70% of that cable plant to support slower-speed two-way data communications, Schryer says. By year-end, AT&T hopes to have 75% of its cable networks upgraded to handle two-way data and voice.
Bringing the other 25% up to snuff could take another two years, says Greg Braden, executive vice president of telephony operations for AT&T Broadband.
Upgrading a network, though, is different from rolling out services. While the infrastructure is ready in many locations, services are lagging because rollouts require installing provisioning gear, launching marketing campaigns and addressing other peripheral issues.
There's much at stake
The grand scheme is not without risk. Wall Street is getting increasingly antsy about the high-cost cable gamble, which is pummeling the company's stock. AT&T said last week that in 10 of 19 cities where it is already offering local voice it is giving the service away to try to win customers.
The greater risk, however, is standing by idly, waiting for local exchange carriers (LEC) to wade into long-distance and start pitching end-to-end services.
Experience shows long-haulers are vulnerable to that LEC sales pitch. Speaking at the recent Vortex conference, William Esrey, chairman and CEO of Sprint Corp., noted that when Southern New England Telephone (now owned by SBC Communications) started selling long-distance in Connecticut and Rhode Island, it captured 35% of the market. "If SBC (Communications Inc.) and Bell Atlantic (Corp.) enter long-distance and walk away with a 35% share, they will have more than MCI WorldCom (Inc.) and Sprint combined," Esrey said.
With the long-distance market worth roughly $100 billion and local markets another $100 billion, each percentage point shift represents an enormous stake.
Another reason AT&T is so frantic to rush new broadband services to market is that consumer research by AT&T Labs shows broadband access is the stickiest telecom service ever offered.
"You are more likely to die than change broadband access," Schryer says.
The implication for AT&T is ominous. If LECs can sneak DSL into homes before AT&T gets its cable act together, it will be hard for AT&T to unseat the incumbent service and realize its cable dreams, to say nothing of holding onto its traditional long-distance business.
Owning the broadband pipe that leads into the home is critical because once the infrastructure is in place, the real money is in selling enhanced services. Schryer says AT&T hopes someday to squeeze $150 to $200 per month out of each home with combined broadband data access, telephony and entertainment services.
Although some of the cable systems AT&T has acquired are modern enough to support such mixed-media services, future high-end offerings will be supported by AT&T's LightWire metropolitan-area network architecture.
Networks built to LightWire specifications are less expensive to run than traditional hybrid fiber coaxial (HFC) networks, flexible enough to scale gradually with demand and rich enough in fiber to satisfy even the most insatiable bandwidth hunger.
LightWire supports a minimum of 2M bit/sec downloads and 200K bit/sec uploads, fast enough to rival most DSL services. By comparison, HFC networks deliver 10M bit/sec per coaxial loop, but the bandwidth is shared by as many as 500 customers, meaning actual throughput is considerably lower.
The main difference between LightWire and HFC networks is that fiber is pushed deeper into the network - closer to customers - by installing fiber multiplexers called MuxNodes and, even further downstream, mini-fiber nodes.
Traffic is pushed to the MuxNodes over wave division multiplexed links, and from there it can be broadcast to all mini-fiber nodes or narrowcast only to nodes that need it.
The architecture offers a number of benefits. There are more opportunities for narrowcasting, which conserves network bandwidth. It limits the length of coaxial cable run to the home to less than about a half-mile, eliminating the need for repeaters on the coaxial lines. That saves AT&T equipment costs and makes the network more reliable because there are fewer electronic components that might fail.
It also means the coaxial cable runs support to as few as 50 homes, says Vice President of Telephony Operations Braden, so there is more coaxial bandwidth to share.
Perhaps most important, LightWire lets AT&T increase network capacity incrementally, according to Sheryl Woodward, a broadband access researcher at AT&T Labs. This can be accomplished in several ways:
As demand for targeted services grows, the company can increase the portion of cable spectrum used to support narrowcast traffic by displacing broadcast channels. This can quadruple the capacity of the network over time.
Taking the narrowcasting one step further, minicasting will let AT&T target traffic to still smaller groups, increasing the useful capacity of the coaxial plant by a factor of eight.
And upstream, AT&T can open currently unused radio frequencies between 900 and 1,000 MHz, boosting capacity 10 times in that direction.
These options together allow for a smooth evolution from LightWire's minimum 2M bit/sec downstream, 200K bit/sec upstream capacity to its full potential of 64M bit/sec downstream and 6M bit/sec upstream, Woodward says.
Because the improvements can be phased in gradually, AT&T can increase capacity when needed instead of having to overbuild and wait for demand to materialize, Woodward says.
But the basic LightWire architecture still involves extra fiber and more fiber nodes, which aren't cheap. As a result, AT&T is only installing LightWire in new installations, such as subdivisions and in networks where wiring is strung from poles, making it easier to install fiber and equipment.
AT&T has no specific services yet in mind for full-blown 64M bit/sec LightWire installations.
Schryer says that much bandwidth will make possible things that have never been dreamed of. Customers in Florida, Utah and Texas are already sending traffic over LightWire infrastructure on a trial basis. The company recently deployed LightWire in Texas to support cable television and data traffic.
There are detractors among industry observers.
"I think it's far too soon to say whether we're going to see 64M bit/sec delivered reliably to each home," says Lisa Pierce, an analyst with Giga Information Group Inc., a Cambridge, Mass., consulting firm. "I think by the time they're done, it's probably going to be more like 10M bit/sec to 15M bit/sec."
Not if Schryer has his way. He says flatly: "We'll eventually do fiber to the home."
AT&T, of course, doesn't need LightWire to support basic voice over cable and is busy installing the voice gear needed to make that service a reality.
By year-end, the networks passing some 75% of the company's cable homes will be capable of carrying voice. However, the company won't say when those users will be able to buy the service.
Migrating voice services
What AT&T will discuss is the plan for migrating voice services from the circuit-switched domain to a full packet environment.
As a first step, AT&T is dedicating the equivalent of two broadcast TV channels to the job of carrying phone calls. This is enough to support hundreds of simultaneous voice calls, Braden says.
Incoming calls are fielded by traditional Class 5 local telephone switches connected to cable TV headends, then dropped on fiber links leading to customer sites. At multiplexer nodes, optical signals are converted to electrical and sent along the coaxial cable to the customer.
At customer sites, a network interface unit strips the voice traffic and diverts it to the existing inside telephone wiring. Customers use their regular phones.
In Phase 2 of the migration, the Class 5 circuit switches will be replaced with packet-based soft switches, which perform the call-handling functions.
In the final phase, the cable telephony standard known as Data Over Cable Interface Specification is adopted and voice is packetized at the handset and transported accordingly.
Schryer says AT&T's voice bid will be aided by the fact that the company's pricing model is different than LECs' pricing.
"The [regional Bell operating companies] lose money on residential voice service and make it up on business. We can drive prices down - mop the floor with them," and make it up on higher-cost, higher-profit interactive services, he says.
Schryer admits that AT&T is in a horse race with the LECs and DSL providers. But he says the company will install 400,000 telephone sets by year-end, and claims the company is installing on a monthly basis more two-way cable customers than all DSL service providers combined.
Whether AT&T's grand cable strategy will pay off is uncertain.
Giga Information's Pierce says, "They're spending tens of billions of dollars and to make that pay for itself they basically have to say, 'I'm going to have the majority of the high-end users.' Well, gee, that's everybody else's plan out there, too."