SDN architectures

Optical SDN Could Require Vendor-Specific Help

NEW YORK -- Packet-Optical Transport Evolution 2013 -- The photonic angle to software-defined networking (SDN) is going to require vendor-specific technologies rather than one industry-standard approach, according to Glenn Wellbrock, the director in charge of Verizon's optical-network architecture.

Wellbrock's talk, which kicked off the POTE conference Tuesday morning, brought a dose of carrier reality to what was certain to be an SDN-heavy day.

His goes against the grain of SDN's promise, where a few, possibly homogeneous controllers rule the network -- but then again, optical transport gear isn't like the rest of the network. The transport layer deals with the physical properties of light, a very analog problem. An SDN controller can't tell, for instance, that a particular span should be avoided because its polarization-mode dispersion is too high.

Moreover, many of the controllable variables in the optical network are vendor-specific. Vendors are trying to stand out by using their own versions of forward error correction, for instance.

Standards are already being hammered out for SDN in the transport layer, but those efforts mostly target the OTN switching layer. Wellbrock was talking about the photonic layer, "Layer 0," as it's called (and that's the term Wellbrock used).

Specifically, he's concerned about reconfigurable optical add-drop multiplexers (ROADMs). For years, Verizon has preached a dream of ROADM networks that are colorless, directionless and contentionless and that run on a flexible grid. All of that freedom would give the operator more options for where to direct wavelengths to go and when those commands could be issued. In turn, that would address concerns that the optical network is too expensive and too slow to configure.

All of a sudden, though Layers 1 through 3 are being blessed with that kind of flexibility, as SDN promises the ability to program network paths and provision services on the fly. Wellbrock's concern is that the optical layer could be left behind as all these other layers improve.

"If they just optimize the upper layers and forget about the physical layer, especially the transport, we're going to be in the same boat two or three years from now, where we ask, 'How do I get it cheaper?'"

The answer doesn't require creating anything new, Wellbrock said. Rather, it's a matter of telling the upper layers what's available at the packet layer: where the possible endpoints are and what bandwidth they can handle, and what the connectivity map connecting them looks like.

But that implementation has to be vendor-specific, Wellbrock argued. One answer would be to use the element management system (EMS). The SDN controller would talk to the EMS, which would make the necessary changes in the ROADMs.

This is all predicated on those "CDC" (or "CDCF") ROADMs becoming a reality, because that would create the optical switching that SDN would control, "but that reality is coming in 2014," Wellbrock said.

As an aside, Wellbrock mentioned that bandwidth-on-demand -- one SDN application that many vendors tend to gravitate toward -- doesn't interest Verizon. Users wouldn't ask for it until emergencies crop up, so Verizon would likely have to build up extra capacity to accommodate the service -- and building excess, unused capacity is exactly the kind of thing Wellbrock is trying to avoid.

For more

— Craig Matsumoto, Managing Editor, Light Reading

fmenard123456 5/18/2013 | 3:18:44 PM
re: Optical SDN Could Require Vendor-Specific Help Optical SDN in the core is only going to become a reality once more granular flows are permitted to transit across the core. This will require Optical Burst Switching with QAM-ish modulations, Automated path orchestration of DWDM waves is for multipoint in the metro.
chechaco 5/15/2013 | 2:01:07 AM
re: Optical SDN Could Require Vendor-Specific Help "An SDN controller can't tell, for
instance, that a particular span should be avoided because its
polarization-mode dispersion is too high."
I think that there's no technical reason why an SDN controller would not be able, in time, understand optical qualities of a path. Of course, if properly explained. I can point to WSON work at IETF's CCAMP WG. Certain parameters and metrics, as defined by ITU-T, are common, standardized. And yet there's plenty of room to characterize vendor-specific parameters and metrics of optical path.
Craig Matsumoto 5/14/2013 | 9:09:22 PM
re: Optical SDN Could Require Vendor-Specific Help The ONF's transport working group is looking at this a different way. The controller can issue abstracted commands that mediation software translates for individual network elements. That way, the controller doesn't have to know all the analog nuances of the elements, (as Mitch Auster of the ONF put it just now in an afternoon POTE session).

That sounds more promising than the EMS model, but it still requires input from each vendor about their machines' specifics. But the ONF's advantage is that they've got an entire group tackling the issue.
Sign In