How Infinera Turned Thinking about Optics on its Head

A Q&A with Jagdeep Singh

Infinera recently unveiled its roadmap for photonic integration, detailing plans to increase capacity per chip from today’s 100gbps to 400gbps by 2009 and eventually to multiple terabits per chip. What will these leaps in capacity entail in terms of innovations and network/product upgrades?
Singh: “There are two main ways to get more capacity per chip. One is to put more bits per wavelength, the other is to put more wavelengths per chip. More bits per wavelength comes down to more sophisticated modulation techniques. For example, the chips that we announced at OFC use a type of phase modulation called DQPSK.

“Another approach would be to put more wavelengths on a chip. Today we have 10 wavelengths on a chip. You can imagine putting 20 or 30 or 40 on a chip. In fact, we did a demo two years ago at OFC that showed a 40-channel chip with 40 gigabits per second per channel.

“The point I’m trying to make is there are two avenues, and we’ll using both avenues.

“The other point I want to make is the industry talks a lot about higher speeds per wavelength, like 40gig per wavelength. It’s important to note that speed for wavelengths is in many ways a completely irrelevant parameter. It doesn’t determine anything that is of interest to the customer – it doesn’t determine the cost of the system, it doesn’t determine the capacity of the system, it doesn’t determine the reliability of the system. All those factors – capacity, reliability, power – are determined by capacity per chip.”

What are service providers’ biggest challenges today in terms of optical networking?
Singh: “There are probably three fundamental issues they’re facing. The first is how do they scale their networks. The good news is bandwidth is growing again because of things like video. But the revenue per bit [the carriers] can command is dropping. That raises a challenge – how do they scale their networks cost effectively? That’s a key requirement.

“The second aspect of it, though, is that even though they know that bandwidth is growing, they don’t know what flavor that bandwidth comes in. In other words, there’s a lot of uncertainty about service types. Is it going to be Ethernet? Is it going to be SONET? Is it going to be point A to B or point A to C, for example. Which cities? How many wavelengths between cities? All of these things are unknowns; the carrier doesn’t know until the customer orders the circuit. Dealing with that uncertainty is a key requirement, but unfortunately the traditional gear, because it’s analog, doesn’t do a very good job of dealing with that uncertainty.

“The final point is in the long run what the carriers need is a way to try to stem the decline of that revenue per bit. They can try to squeeze their vendors on cost, until they get to the point where the revenue per bit declines and starts to flatten out and they’re going to be under pressure from a business model standpoint. So they need equipment that will help them deliver differentiated service, so they’re not competing with their competitors with the exact same product, which will result in purely price competition.”

Share this article: Email, Slashdot, Digg, Del.icio.us, Yahoo!MyWeb, Windows Live Favorites, Furl
Add this article feed to: RSS, My Yahoo, Newsgator, Bloglines