Pieter Poll

SAN DIEGO--Pieter Poll, Qwest Communications’ chief technology officer, described some of the innovations critical to the future of optical networks, but he also identified hurdles to the commercial implementation of those innovations in a speech at the OFC NFOEC show today.

Qwest is seeing demand today for 40 Gb/s links today, Poll said, but the economics of the technology are still problematic. Though carriers complained a year ago at this show that the cost of 40 Gb/s needed to come down, those costs are still too high, he said. “The cost advantage on the IP side drives us to use n-by-10 GigE meshes versus 40 Gb/s pipes.”

And he hinted that migrating directly to 100 Gb/s technology would continue the cost benefits the telecom world saw migrating to 10 Gb/s and enjoying the economies of scale seen in the 10-Gb/s enterprise networking world.

“At Qwest, we’re not keen to see those worlds separate,” Poll said. “We’d like telecom technologies to continue to track enterprise technologies in powers of ten.”

However, he acknowledged challenges in getting to 100-Gb/s and pointed out that some technologies introduced to enable 40-Gb/s migration, such as multilevel modulation schemes, might also aid 100-Gb/s migrations. “Leveraging 40-Gb/s technologies may make this transition simpler than from 10 Gb/s to 40 Gb/s.”

Vendors have told Qwest that commercially deployable 100-Gb/s telecom gear could arrive by 2010 or 2012, Poll said. “Certainly we’d like to see it sooner.”

Poll called G.709 Optical Transport Networking technology “critical” for Qwest due to its ability to transport 10-Gb/s signals across the network.

On another subject, he also lamented the challenges in implementing optical control planes that would allow for dynamic large-scale provisioning.

“The benefit so far of a true optical control plane is somewhat minimal,” Poll said. “The ability to move wavelengths around dynamically is still operationally some ways away.”

An optical control plane would theoretically speed up provisioning and increase flexibility in restoration moves. But the realities of today’s networks make the realization of those benefits difficult, Poll said. Signals can’t simply and immediately be rerouted from, say, a southern cross-country route to a northern one, he said, owing to the unique characteristics of each route; different paths harbor different distances, different fiber types and equipment from different vendors.

Though networking trends point toward an evolution to mesh networks, nationwide carrier networks currently lack the physical diversity that would help carriers realize the benefits of true mesh networking, Poll said. Qwest, for example, has about three or four cross-country arteries that correspond to railway rights of way. Replacing that with a more mesh-like architecture would increase the complexity of operating the network. For one thing, it would require more uniformity in the capacities of various network routes.

“You’d have to have units of 10 Gb/s traffic between all points on the network before this becomes economically viable,” Poll said. “When you place IP capacity, you have to place a lot of standby capacity to carry traffic along different paths. If we could get greater physical diversity in place, we could greatly diminish the amount of standby capacity we have to take.”

In order to realize the benefits of mesh networking, Poll said, carriers will need to cooperate with each other more than they currently do, using fiber swaps to increase the geographic diversity of network paths.

Like some other large carriers such as AT&T and XO Communications, Qwest has determined that IP-over-DWDM, an approach that integrates long-haul optics in IP routers, is not a cost-effective choice for it. Although the approach eliminates interfaces between optical gear and routing, it also introduces other challenges that impact the overall economic benefits.

“It boils down to straight cost,” Poll said about why Qwest chooses not to use IPoDWDM. Comparing it to a 10 Gigabit Ethernet LAN PHY interface between long-haul optics, “The economics overwhelming favor doing the two next to each other, lying short-reach interfaces from the data supplier [with] long-haul [gear] from the optical supplier.”

An IPoDWDM approach also brings with it “a bunch of management issues,” Poll said, because it introduces so-called “alien wavelengths” whose performance is obscured to some degree from the carrier. Integrating IP and optics is also an operational challenge for large carriers because those functions are today typically managed in separate locations by teams with separate expertises.

Before closing, Poll urged the optical industry to work harder to develop all-optical networking as an alternative to routers as a way of keeping up with booming traffic trends.

“The terabit-class routers today, you’ll dry your hair from the exhaust that comes off those devices,” he said, citing 10 kilowatts per rack as typical.

“Optical buffering needs to take place or synchronization across networks at the level of flow, with fixed time slots,” he said. “Simple networks like this have already been shown to work in research labs but are not realistic in the near future.”

As one possible remedy, he suggested very fast tunability at the wavelength level, using existing technologies to quickly separate packets based on wavelengths. (Poll’s description resembles the approach of Matisse Networks, though he didn’t mention that vendor by name.) “This is achievable much earlier than true end-to-end optical packet switching,” he said.