In the Spotlight: Drew Perkins, Infinera

Infinera last month trumpeted its ascension to the top of the global 10-Gb/s long-haul optical systems market, as revealed in a recent Dell'Oro report that attributed 20% of the market to Infinera in the third quarter. The news didn't sit well with Philippe Morin, general manager of Nortel Networks' optical business, who said Infinera was selling a lot of 10 Gb/s ports because it was crowding its customers' networks with regenerators every 80 km, while Nortel was aiming to stretch its products' reach. Infinera's founder and Chief Technology Officer Drew Perkins fired back, adding his own insights on the near future of the optical sector.

On Morin's comments: I think Philippe is trying to spread fear, uncertainty and doubt. That's poppycock. We enable customers to have add/drops every 80 km or less because we can cost-effectively do that, whereas most vendors cannot. We regularly have customers deploying it at kilometers and tens of kilometers between add/drop sites. However, we have customers that deploy our equipment 500 or 600 km between regen[erator]s or add/drops or any electronic sites. We support normal EDFA-based optical amplifiers just like everybody else. Carriers don't want to be able to drop services every few hundred kilometers; they want to be able to drop services everywhere there's a customer. Profitable customers turn up in small cities as well as large cities. The more profitable customers are the ones in the smaller cities only tens or hundreds of kilometers away, not 500 km away. You need equipment that can cost-effectively allow you to drop traffic very frequently so you can deliver those services so you can build your network around the services you want, rather than offer services around the network equipment that you have.

On integrating IP and optics: It makes a great deal of sense to strengthen the communication between IP routers and the optical network to allow much greater optimization and save capex and opex. There's a lot of standards activity around the integration of the IP layer with the optical transport layer. Those initiatives started first in the Optical Interworking Forum by myself many years ago when I was vice chairman of the OIF technical committee. What there isn't right now is any cohesive group of suppliers and customers, such as an alliance. It may make sense to have such a thing. We're interested in feeding into that process as well, to gather even more customer feedback and get more requirements feeding into the standards groups that are working in this area. An alliance could help clarify a set of requirements going into the standards groups, increase the number of carriers getting involved and speed things up by pushing along the standards groups. There are already a substantial number of carrier service providers that are involved in the OIF, the IETF and the ITU. But there are a lot that are still not.

On optical scalability: For years the optical world has talked about how scalable [wavelength-division multiplexing] optics was because you could increase the performance by adding channels. But they were completely missing the point. The other part of Moore's Law is that increases in performance are done at a constant cost. It's not the performance that increases; it's the performance per dollar. Until the photonic integration that we've introduced, that part has been completely missing. It's not nearly as powerful for optics to continue to increase performance if it does so at a similar increase in cost. It's all about how are we going to get hundreds of gigabits, terabits, tens of terabits [per second] in the not-too-distant future but do it at today's costs.

On 40 Gb/s and 100-Gb/s wavelengths: I can't pin down a particular date for delivery of 40-Gb/s wavelengths. I can't pre-announce products like that. [Photonic integrated circuits] have incredible scalability. They scale with a Moore's Law-like process. Doing a 400-Gb/s PIC [presented in a paper at this year's OFC/NFOEC show] was extremely straightforward--simply turning the crank on a 100-Gb/s PIC. Most of the industry has linked the optical line system with the service delivered. That linkage really is irrelevant to carriers. Carriers sell services, not wavelengths. If they're selling 40 Gb/s services, it really doesn't matter if that's being provided as a 40 Gb/s wavelength or a 10 Gb/s wavelength. In fact, it's very unlikely anytime in the future that carriers will be able to deploy a 40 Gb/s wavelength infrastructure end-to-end. They'll need to provide 40 Gb/s services, even over portions of their network that are not 40 Gb/s-capable. For the people who are linking the bandwidth of the service with the bandwidth of the line optical system infrastructure, how are they going to do 100 Gb/s services?