Glenn Wellbrock has been working with 40 Gb/s networks for years. At MCI, he helped conduct trials of 40 Gb/s networks starting in 2004, anticipating a cost benefit from 40 Gb/s migration. But last month, as director of network technology development for Verizon Business, he announced the company’s first commercial 40 Gb/s deployments, saying it was necessary to relieve capacity constraints in some places but that the cost benefits aren’t there. He recently spoke with Telephony’s Ed Gubbins about deploying 40 Gb/s and 100 Gb/s networks, integrating optical and IP networks and breaking down the wall between routing guys and transport guys.

On 40 Gb/s deployment: Most routes have at least 40 Gb/s [of traffic] on them, but it isn’t necessarily cost-effective to use 40 Gb/s everywhere yet. It’s all up to the equipment manufacturers. If they lower the cost, we’d use more of it. But until then, we’re inclined to use more 10 Gb/s channels like we do today. We will use 40 Gb/s where it’s needed. On these high-congestion routes where it makes sense, of course we’ll do it. We’re not saying whose gear it is. There are multiple vendors involved.

On the 40-Gb/s problem: [The cost of 40 Gb/s compared to four 10 Gb/s channels] is not good enough. Our target was [no more than] 2 to 2.5 times the cost, and we’re not there yet. Everybody keeps complaining. The component guys are complaining, the systems guys are complaining they don’t have enough volume, so the parts aren’t there. It’s different than Ethernet rates, so it’s not leveraging the Ethernet ecosystem--all the components that would come out of Asian countries and such. We think 100 Gb/s would help solve that. If telecom becomes the same data rate as Ethernet, as the IEEE and the ITU-T agree on it, we’d be able to leverage chip sets and everything that are the same data rate we use for transport. We think that will help. It took a long time to get to 40 Gb/s. We had to do a whole lot of pretty sophisticated new innovation. Going from just the on-off key in the optical sense to different modulation formats, even using different receiver types. Electronic dispersion compensation, optical dispersion compensation. We’ve really done a lot of innovation to try to make 40 Gb/s viable for ultralong-haul distances. That took quite a while. And the sophisticated modulation formats are all brand new. This is the first time we’ve ever used them. If we apply those to 100 Gb/s, we get a 10x instead of just a 4x difference in capacity from 10 Gb/s. And if the cost is similar because we’re using parts similar to 40 Gb/s, then that’s a big bargain for both us and the systems guys.

On hopes for 40 Gb/s: The cost of 40 Gb/s was always deemed, “Well ITU-T standardized on it; maybe that will be enough.” And if we can get four times ten, maybe that will be enough to get it cost-effectively. We couldn’t do any more than that really. But our hope of going to 40 Gb/s was that it would be cheaper than four 10 Gb/s [links]. Certainly more spectral efficiency. You’ve got 80 wavelengths to work with on most systems, 128 on others, 160 on others. With that number of wavelengths and higher data rates, you get more throughput, better spectral efficiency. There’s always value from having fewer, bigger pipes. From carriers’ point of view, if we can carry everything on one wavelength, that’s the way we’d do it. Fewer things to manage, fewer things to spare, to repair, to troubleshoot, to do network management for--the whole thing is simpler for us. But can we pay a premium for it? Not really. We still have a lot of competitors out there fighting us on every bid.

On integrating optical and IP: It reduces a transponder. The back-to-back portion of that, the gray interface [vendors such as Cisco Systems] are eliminating, is not the most expensive part. So it’s a marginal cost savings, but it is beneficial in that it’s fewer things to manage and spare. But today those transport systems aren’t Cisco’s or Juniper’s. And yet those are the two big switch router guys. All the switch router guys don’t necessarily own the transport gear. So then you have to operate it in what’s called an alien wavelength, which means the transport system is just accepting a signal it knows nothing about necessarily. It can do power management on it, but it can’t do the same type of controls it would have over its own wavelength, so the performance is a bit limited. Especially in a metro environment, it can be done quite well. It could even work for long-haul possibly. If the router guys really do integrate the best technology into that transponder and we can make it work with our currently deployed transport systems, then it’s a workable solution. But the router guys really have to make it advantageous to do that. If they’re charging a premium for it, then all they’ve saved is that gray interface, the 1310 nanometer short reach between the two devices. They have to make it a lower total cost than it would be to combine their 10 Gb/s Ethernet port today. This is part of the other problem: If you wrap it up in an [order 2 optical data unit (ODU2)] wrapper for 10 Gb/s anyway, you’ve got to add forward error correction, you’ve got to have an ODU2-compliant signal that enters the transport network, but you have to do it at the same cost the transport guys are doing it. Because today the trade-off is a 10 GigE port, and that can even be a LAN PHY if you want, which is inexpensive, going over to a transport 10 Gb/s pipe. So we’ve got to look at that whole picture. They’ve got to make it cost-effective for us to do it that way. They can’t say, “Well, I’m dragging that over into the router, and it’s going to cost you a premium.” It has to be a lower cost to the carrier than putting it in the transport gear.

On the cultural wall between transport guys and router guys in carrier ranks: They are managed by different teams. Will we overcome that stuff as carriers? Yeah. It’s a matter of survival and being competitive. Convergence is changing a lot of the way we do business. It will evolve to where most of the hardware isn’t necessarily completely different people. A lot of the services, software and back office--those are the primary differences. Many times, field technicians are pretty adaptable. These guys can work on anything. Most of the time it’s the IP services versus the transport services and how they’re handled in tech support centers and back office systems--these are the big issues.

On capex versus opex savings: A lot of times, we address the opex portion of it by having fewer interfaces to manage and fewer boxes. That drives the opex savings. But it’s hard to trade that against the upfront capital cost. It’s hard to deploy things just because there’s an opex savings in the end. If it’s lower capex and saves opex, it’s a no-brainer. Today LAN PHY interfaces are pretty reasonable compared with the long-haul ones, plus a transponder from companies who primarily build them.

On photonic switching: Photonic switching is something we’re seeing finally coming along. How do we manage automated fiber patch panels? Patch panels are one of the things we don’t have visibility into in the network. We’ve got large switches and routers--we can see everything in there. We have transport gear with control planes. And we’ve got a patch cord between them. In the outside plant (OSP), we’ve got control planes built right into the transport gear. It’s where we’re headed anyway. And now we’ve got a patch cord that runs to the OSP. That’s a part of the network we don’t have visibility to or control of from a remote system.