The Photonic Pioneer

Level 3 Communications' widescale deployment of Infinera's DTN optical switch, announced in late May and commencing in the third quarter, is both a resounding validation and the ultimate test of a unique new transport technology that analysts have suggested could represent a profound advancement in network technology: integrated photonics.

To meet the demand for increasing capacity in its IP backbone network, Level 3 is deploying Infinera's DTN across “the vast majority of [its] footprint” in the U.S. and Europe, according to Randy Dunbar, Level 3's senior vice president of transport services. That's some 23,000 miles of intercity fiber. The carrier expects the domestic network upgrade to be complete by the end of the year.

Level 3 has had test routes running with the DTN for months. The carrier initially evaluated dense wavelength division multiplexing (DWDM) platforms from more than 10 different vendors, including all incumbent suppliers, and chose Infinera's based on its high performance across a number of categories, including scalability, operational efficiency, cost and the timetable for offering new (i.e., even higher capacity) services.

“We saw other platforms that would do well on one of those measures,” Dunbar said, but Infinera scored the highest across the board.

The amount of traffic on Level 3's IP backbone, which rides atop the company's DWDM network, is doubling annually, Dunbar said. Demand for more capacity is coming from all directions, including domestic and international carriers building IP and data networks and enterprises looking for disaster recovery. Level 3 saw a need for 2.5 Gb/s and 10 Gb/s services and, perhaps next year or so, 40 Gb/s services.

“At the rate we're growing, we need to be able to scale the network extremely fast,” Dunbar said. “This is the kind of investment we're making to make sure we have all the capacity we need to keep up with that growth rate.”

For Level 3, the deployment is a hefty bet on a bold new technology and a chance to scale network capacity using a vastly more amenable cost structure than previous technologies offered. For Infinera's founders, it is the test they needed to prove their claims.

“What [Infinera] won was the opportunity to prove their concept,” said Dave Dunphy, analyst with Current Analysis. “Obviously, some people who saw it in the lab thought it was reliable or it wouldn't have scored the kind of win it did.”

Naturally, the industry has been keeping an eye out since then for real-world deployment of the DTN. In January this year, Infinera announced its first customer: German ISP FreeNet. But the really big news would come four months later with a major deployment commitment from a carrier spanning two continents.

As a group, analysts have been hard pressed to point out potential drawbacks or downsides to Infinera's approach. In fact, the primary misgiving analysts had was about whether the start-up's gear could actually do what its founders claimed.

“It's a problem so difficult, most design teams haven't even tried it,” Infonetics Research founder Michael Howard said when Infinera exited stealth mode just more than a year ago. If Infinera's solution works, it could be “a major breakthrough” and “a disruptive technology,” he said.

Founded in 2001 and well-funded with more than $200 million by last fall, Infinera tackled a problem that had been addressed years earlier by equipment vendors such as Corvis, Calient Networks, Xros Communications and others: namely, the high cost of the optical components that are used to convert optical signals into electronics and back, allowing carriers to monitor, switch and otherwise manage traffic while in electronic form. Vendors of all-optical gear addressed the issue by omitting those conversions altogether, using a variety of methods to bend and route lightwaves. Though all-optical networks met with some success, they didn't revolutionize telecom networks quite the way they promised.

Infinera took the opposite approach. Rather than get rid of those expensive optical-electrical-optical (OEO) conversions, Infinera tried to simply make them much less expensive. In the DTN, the OEO conversions that normally would be performed by dozens of discrete photonic components are done in a pair of 100 Gb/s indium phosphide chips that Infinera designed. Each transmitting chip does the work of 10 lasers, 10 modulators and a 10-channel multiplexer; each receiving chip contains a 10-channel demux and 10 photo-detectors. The pitch, which Infinera calls “digital optical networks,” is a best-of-both-worlds promise: all the transparency and manageability of electrical switching without the high cost of OEO conversion.

“The only downside that I can see [to Infinera's approach] is a religious argument,” said Doug Green, principal consultant with the Bradam Group. “Some service provider CTOs have been preaching all-optical so long that they may have a hard time swallowing their words.”

One of the most impressive features of Infinera's approach, according to Level 3, was its scalability. Each half-rack chassis can handle 400 Gb/s or scale to 800 Gb/s in a full rack. Each line card includes two photonic circuits and yields 100 Gb/s.

“Every time you plug in a new line card, it's like adding another 10-channel WDM system,” said Rick Dodd, Infinera's director of marketing.

That allows Level 3's network planners to err on the side of having excess capacity rather than having to add another transponder card and light another wavelength every time they want to provision another 10 Gb/s. And they can turn up links quickly because they don't have to tune or adjust optical components. The DTN's muxponder cards can also subdivide each 10 Gb/s link into four 2.5 Gb/s links. Level 3 generally runs its IP backbone on 10 Gb/s wavelengths but sells waves in 2.5 Gb/s and 10 Gb/s sizes.

The DTN's digital nature allows carriers to monitor and manage traffic at each node in the network rather than just the endpoints, as they might in WDM networks. The DTN provides per-circuit bit error rates at every node, both before and after forward error correction. So, for example, if a splice in the network is causing degradation, a network operator would be able to see bit error rates before they're corrected, allowing him to pinpoint the problem and fix it before the customer sees any bit error.

“It's very different from conventional WDM, where I may be able to see bit error rates only at the end points of the network, where I have to infer the quality of service based on some analog parameter like the optical signal-to-noise ratio or the optical power received,” Infinera's Dodd said. “Your network becomes highly manageable because you have great digital information at all nodes and on all services in the network.”

Also, each DTN includes GMPLS routing and signaling, so each switch is aware of the overall network and how to light capacity quickly between end points.

Level 3 hopes its bold choice of technology will not only fulfill its capacity needs in an efficient way but also act as a differentiator going forward.

“A lot of our competitors have been selling excess capacity, but when they run out of capacity, they're not willing to continue growing or augmenting their network at the prices in the market today,” Dunbar said. “Many of the traditional carriers that have announced next-gen deployments have announced traditional suppliers.”

However, assuming the Level 3 deployment goes well, other large carriers across the globe are sure to start paying more serious attention to Infinera, if they aren't already.

Meanwhile, the vendor isn't done pushing the scalability ceiling. At this year's Optical Fiber Communications/National Fiber Optics Engineers Conference in March, only 18 months after introducing 100 Gb/s chips, Infinera's founders presented a paper describing the use of 400 Gb/s chips. To the Infinera team — a knowledgeable group that includes veterans of Lightera Networks, the company that would eventually yield, through acquisition, Ciena's CoreDirector optical switch — one of the only things left to be proven is exactly how far the technology can go.

“We think there's going to be rapid improvements in the capacity and capability of photonic circuits,” Dodd said. “In the same way that, with Moore's Law, we saw very rapid increases in the capacity and capability of electronic chips, it's going to be exactly the same in photonics — very rapid innovation.”