Breaking the Metro Bottleneck
The information racing across the country in huge fiber-optic pipes hits a snarl under city streets. New optical networking techniques are clearing the way.
In the so-called backbone of the telecommunications system, the fat pipes that pour data across continents, the name of the game is raw speed (see “Building a Better Backbone,”). But the data racing through the telecom backbone can’t fulfill its mission until it is shuttled through the “metropolitan loop,” a complex network of cables and switches that delivers those bits to businesses, factories, schools and homes. It’s there that the information gusher narrows to a relative trickle, because the metro loop is every bit as tangled as downtown rush-hour traffic. If the broadband revolution is ever to be a reality, the metropolitan bottleneck must be broken.
But that’s a tall order. Upgrades in the metro loop have been far slower in coming than advances in the backbone. The reasons range from tougher cost constraints to urban bureaucracy to the presence of a patchwork telecom infrastructure dating back to the 1970s and ’80s. But R&D aimed specifically at the metro loop is slowly pushing a variety of solutions out of the laboratory and under the streets. And if we really want broadband, those fixes had better work.
To grasp the scale of the bottleneck, consider the metro network’s place in the telecom ecology. In the backbone, transmission speeds are measured in trillions of bits per second. On the user end, high-speed networks run at billions of bits per second (gigabits). But the metro systems that link these two high-speed networks poke along at mere millions of bits per second (megabits). “That’s the bottleneck,” laments Steve Schilling, presi-dent of access networks at Nortel Networks. And this metro-loop constriction isn’t just a problem for the companies that run the networks. Regular folk experience it as trunk-line busy signals and stalled Web browsers.
If you’re looking for a culprit here, don’t finger the phone companies that run the metro loop. They planned prudently (at least, so they thought) for steady growth in voice communications, which at the time was their bread and butter. Then they, along with everyone else, were blindsided by the explosion of the Net. “Two to three years ago, we started running into capacity problems” in urban areas, says Stuart Elby, who heads development of Internet-connected networks at Verizon, the phone company serving New York and New England. Speeds of 2.5 gigabits per second, enough to handle heavy Net-generated traffic, are common only in the hearts of big cities like New York or Boston, where Verizon runs fiber-optic cable with 48 strands. More typical metro-loop speeds range from 1.5 to 600 or so megabits per second.
And there’s no letup in sight for the beleaguered metro operations. “More and more applications are emerging as you have more bandwidth,” says Claude Romans, an analyst with the South San Francisco market research firm RHK. If digital television ever gets off the ground, for instance, it could gobble up huge chunks of bandwidth; it takes 1.5 gigabits per second to transmit a single, studio-quality high-definition video channel (although consumers will see only a compressed 20-megabit-per-second version). That kind of data onslaught will bring the metro loop to its knees without significant technological upgrades.
The current and future transmission slowdown afflicts both main components of the metro loop’s hub-and-spoke structure. The “access” portion of the network-the spokes-ferries signals out to residential neighborhoods and individual office buildings. These access lines connect to the “collection ring,” which transports signals around a metropolitan area, linking telephone company service centers and other major traffic centers, such as Internet service providers and large universities.
Technological advances are helping unsnarl both the collection ring and the access lines. Fiber optics, which already dominate the collection ring, are replacing more and more of the residual copper in the access lines as well-in effect, paving over dirt paths with smooth, modern asphalt. And new optical transmission technologies are stuffing more data into the networks that are already in place.
Packing Bits and Wavelengths
The heaviest lifting in a metro system is typically done by the collection ring, which runs all the way around the region, providing local access as it goes. To burst through the bandwidth bottleneck here, engineers have two basic choices: they can crank up the bit rate on a single beam of light traveling through a fiber, or they can multiply capacity by using several wavelengths as information carriers. In the second alternative, known as “wavelength division multiplexing,” each fiber carries multiple light beams of different colors-with a different digital signal encoded in each beam. The more wavelengths you can pack in, the more information you move. (These “colors” are actually different shades of infrared and are invisible to the eye.)
Both approaches are now being tried by the companies that run the metro loop. Various technical problems make it tough to raise the bit rate. But in encouraging recent developments, two optical-networking leaders-Ciena and Nortel Networks-have demonstrated single-wavelength transmission of 40 gigabits per second over lengths of fiber typical of a metro network. That’s a big leap over the 2.5 gigabits per second at which today’s fastest metro networks operate. Taking this research feat out of the lab and under the streets, however, will require advances in the electronics that manipulate the signals, since standard chips don’t yet operate that fast.
A trip around the metro collection ring shows that it is stuffed full of fiber; copper has been almost banished. But in the access lines at the fringe of the network-the links that connect the ring to homes and businesses-fiber still coexists with its old-fashioned counterpart. “Fiber is getting further into the access network every day, but it’s got a long way to go,” says Brian McFadden, president of photonics networks at Nortel Networks.
That’s understandable. Even though fiber is cheaper to operate and more stable than copper, established companies can’t afford to rip out all of their installed cables at once. “The amount of infrastructure is enormous; even changing out a few percent a year is a tremendous investment,” says Verizon’s Elby. That’s why a consortium of telecom equipment makers and service providers is pushing to develop evolutionary pathways to bring fiber ever closer to the homes and offices that use the network.
The key technology in this evolution, called a passive optical network, extends the reach of fiber optics further out to the fringes. In order for this technique to work, at least some fiber service must already be in place; but passive optical brings fiber to parts of the network now served only by copper.
Here’s how passive optical works. A transmitter at a central facility generates an optical signal at one of two standard telephone-system data rates-155 or 622 megabits per second. This signal is a composite, which includes information for as many as 32 users. A “passive” optical coupler-which requires no electrical power-then divides this signal among fibers that link directly to end users or to other branching points. Equipment at the end of each of those fibers sorts out the signals, relaying only the ones in-tended for the local user. The central transmitter can reallocate bandwidth among customers almost instantaneously.
For a telephone company, passive optical networking offers an attractive way to extend the reach of optical fiber with minimal fuss. The passive design keeps hardware, operation and installation costs down. Moreover, the sensitive equipment needed to transmit, receive and reroute optical signals is kept safe inside buildings at the ends of the system. And since the passive optical network requires no electrical power between its end points, it generally needs less maintenance than networks based on active components.
A dark-horse technology that has recently joined the metro network, called Gigabit Ethernet, ups the speed ante even further. These systems use fibers to transmit information in the Ethernet format commonly used for office computer networks. Their data rates of one gigabit per second leave other access-line technologies in the dust. A gigabit is 1,000 megabits; gigabit-per-second transmission would, for instance, whisk away the entire contents of a CD in less than a second.
In a Gigabit Ethernet, a single fiber pipeline goes to a central switching point. This Ethernet aggregator, as it is called, distributes signals out to as many as 200 fibers. Each output fiber-like the input fiber-can carry up to one gigabit per second for short bursts, but the total output speed cannot exceed the input. An aggregator box the size of a telephone booth can serve more than 200 homes within a radius of up to 10 kilometers. That’s well beyond the reach of digital subscriber lines, or DSL-the phone company service that provides broadband connections through copper cable.
Gigabit Ethernet can work as a cheap end run around telephone companies for delivery of broadband access. That’s why an Ottawa, Ontario-based nonprofit consortium of companies and universities called Canarie is promoting the technology for broadband connections to cash-strapped schools. In the United States, Veradale, WA-based startup World Wide Packets has developed its own version of the technology for rural telecommunications. It is field-testing a system in Ephrata, WA, for the Grant County Public Utility District.
Merely weaving more fiber into the metro network won’t solve all the problems that crop up in urban areas. Today’s systems rely on a sometimes awkward mix of electronic and optical technology. Tiny lasers launch data-bearing light beams into optical fibers. At the other end, the light strikes a photosensor, which converts the on and off flashes into an electrical signal that electronic switches direct to its proper destination. Such electronic switching works fine at the modest speeds of 2.5 gigabits per second that are now in common metro use.
But start cranking up the data rate, and electronic circuitry has a tough time keeping up with the potential of optical networking. The solution: all-optical switches that redirect light signals without converting them to electrons. The higher the bit rate, the greater the all-optical advantage. Indeed, when you get to 40 gigabits per second, “there’s no alternative to all-optical” switching, says Lawrence Gasman, president of Communications Industry Researchers.
Getting to an all-optical metro system isn’t going to be simple, though, because it will require the construction of new networks. For established phone companies, the burden may not be crushing, since most existing underground urban cables are threaded through buried ducts, and phone companies can often pull out old cables and pull in new ones-as they did when replacing copper with fiber cables in the 1980s. New companies, on the other hand, have to build complete new networks. One such company, Metromedia Fiber Network, plans to expand beyond its New York City base and install almost six million kilometers of fiber in 67 cities in North America and Europe by 2004.
But whether they are laying entirely new networks or trying to modify existing systems to upgrade their performance, the builders and operators of the metro loops that knit together the most concentrated populations of homes are performing a crucial task. The backbone and the corporate networks that flank the metro loop get faster every year. If the metro bottleneck is not broken, broadband will remain little more than a clever idea that some techies once had.
Become an MIT Technology Review Insider for in-depth analysis and unparalleled perspective.Subscribe today