OFC 2000: Optical Switching Systems Stir Up Wavelength Management Debate
BALTIMORE—Among the multitude of optical devices, electrical control chips, and transmission and switching systems on display at the Conference on Optical Fiber Communication (OFC 2000, March 5-10), optical cross-connect systems are gleaning the most attention—both exuberant and skeptical.
Start-up Xros and venerable test equipment house <%=company%> are both unveiling optical cross-connect systems. Lucent also has a system in the works, using switching technology similar to Xros, micro-electromechanical systems (MEMS), basically hinged micro-mirrors, to switch wavelengths. All-optical switching technology, a holy-grail topic discussed at OFC for years, is generating much of the excitement at the event this year as it approaches commercial reality.
Manufacturers of commercially available electrical systems are defending their turf as the only full-function systems available now and extolling their ability to adopt optical switch fabrics when it's prudent to do so. Proponents of big routers continue to make their case that their devices can do the cross-connect job. And novel transport systems circumvent the need for complex wavelength management in the intermediate term by providing bandwidth to spare.
All-optical proposition
A briefing on the Xros device addresses almost all the desired features of wavelength management: Scale (1152x1152 ports), density (three telecom bays), modularity, transparency (10 Gb/s and higher per-port wavelength accommodation), fast protection switching (50 msec), rapid and reconfigurable provisioning, low power consumption (3,000 W), and cost that scales at lower multiples than electronic devices. The system will also feature both user network interfaces and network-to-network interfaces, issues that are being hashed out in technical forums to intelligently link layer three and access equipment to optical core equipment.
Xros' breed of micro mirrors uses single-crystal silicon, which provides elasticity for reliability and thermal properties desirable for handling high-bit-rate signals—including those laden with multiplexed channels, which provides an advantage over other mirror approaches, explains Rajiv Ramaswami, Xros vice president of systems architecture. "We do everything an electrical cross-connect can do, but better," he gushes. Xros is demonstrating a working prototype partially loaded with 24 active ports at OFC.
All-optical future, electrical capability now
Electrical switch proponents are quick to point out weaknesses in the all-optical story. Wavelength management requires intelligence, performance monitoring, multiplexing, and demultiplexing, points out Tellium's Nick DeVito. "One way or the other, you have to pay for those functions," he says, maintaining that the only alternative for doing that right now is with electronics.
Electrical design does not preclude migrating to optics in the future using the same platform. "Our design accommodates o-e-o on the way in, with an all-optical switch core," DeVito says. Without the electrical-to-optical conversion, he says, you can't do wavelength conversion, a critical element for non-blocking switches that can route any wavelength to any other wavelength.
Cisco Systems' Joe Bass, an executive from the Monterey cross-connect legacy at the company and former Alcatel veteran, still makes an impassioned case for the Monterey devices, which are similar to Tellium's. "Monterey wasn't founded on a gee-whiz technology looking for a problem," he says, adding that the devices were designed to solve problems in networks, not push novel technology into application. Like Tellium, Bass' group is open to using attractive new switch fabrics. "If we see a switch fabric we like, we'll use it," he says.
For now, Bass maintains that at least until routers develop robust multi-protocol label switching protocols (MPLS) or optical cross connects develop performance monitoring and wavelength conversion techniques, the electrical cross-connect fabric is the best way to manage wavelengths in the core network, for those networks with a complex array of wavelengths to manage.
Multiple cross-connect bets
Playing all the angles, Lucent has a bet on several cross connect strategies. Its Bandwidth Manager handles 2.5 Gb/s and 10 Gb/s signals with STS-1 granularity. Lucent forged a deal with Tellium to supply customers with Tellium's electrical system that manages signals at 2.5 Gb/s and 10 Gb/s (with 2.5 Gb/s granularity), without breaking them down to STS-1. On the optical side, Lucent is developing the Lambda Router, which handles wavelengths transparently.
"With the wavelength router, because the switch fabric is optical, it will switch 10 Gb/s and 2.5 Gb/s with one port," notes Rod Alferness, CTO of Lucent Technologies Bell Labs' optical networking group. With 80-wavelength DWDM systems being installed, Alferness says single-port 10 Gb/s wavelength management devices are the prudent way to go. "That's a level you want to manage the network," he says.
IP clouds cross-connect future
As if the variety of cross-connect options were not enough, hardcore Internet protocol proponents still maintain that the devices are undesirable. Despite Bass' pragmatic view of core wavelength management, Cisco still maintains a non cross-connect bias. "If you're in the business of selling circuits, these systems are indispensable," says Larry McAdams, product manager for emerging technologies in Cisco's Optical Internetworking business group. But if a carrier's business is selling IP-based services, he suggests that they can skip cross-connects and just use routers.
Bass concurs that routers can adequately handle IP-centric optical network traffic, to a point. As the network scales, he says that an all-IP carrier will need to implement either cross-connects or routers with robust MPLS technology to engineer traffic. As routers scale in capacity to 10 Tb/s, McAdams maintains that routers will incorporate all the good things that cross-connects offer. "What can optics do for us better and cheaper than electronics" is what he suggests as the guiding principle for systems development, noting that Moore's law is hard to beat, and that electronics continue to improve.
Bandwidth alleviator
One way to circumvent the complex array of cross-connect options, at least in the near term, is to take a transport approach like that of Qtera, according to Qtera president and CEO Fahri Diner. The company quit teasing the world about it's technology at OFC by demonstrating a 56-channel, 560 Gb/s system transporting traffic over 3,600 km with no regenerators. Qtera uses a proprietary out-of-band forward error correction technique, distributed amplification (Raman amps), and nonlinear return-to-zero modulation (a quasi-soliton approach). The Raman and RZ accomplishments are firsts, Diner claims.
"I have limited cross-connect functionality today," Diner says of the Qtera system, which performs simple add/drop at network nodes. "You can always trade off complexity for bandwidth." Diner bets that a 560 Gb/s transport platform will accommodate wavelength management needs with add/drop and protection done at the optical layer. "We did this because non-blocking cross-connect technology is not available," he says. "We're evolving."
"People build all of these complex architectures to use bandwidth more efficiently because bandwidth has been expensive and not available," Diner says, referring to the current generation of cross-connects that use electrical methods to do wavelength conversion and other key functions. "If bandwidth is cheap and available, you don't need the complexity.
Ultimately, when network demand requires a hundred times more capacity than Qtera's platform handles now, carriers will once again need to use wavelengths more efficiently, Diner predicts. He envisions a core network of terabit routers, optial cross-connects, and transport systems like Qtera's. But for now, he reasons that Qtera's platform makes up for wavelength routing capability by providing enough bandwidth to make up for it.
Optical cross-connect technology doesn't necessarily need to function within a traditional cross-connect platform, Diner says. "IP routers can do a lot of the stuff optical cross-connects can do," he says. "If you can get non-blocking optical cross-connect matrices with terabit routers, then that's all you need." From there, the carrier connects the router ports to "pipe-provisioning devices" like the Qtera transport system, and the core network is done. At least until the network scales to such complexity that, again, a stand-alone cross-connect becomes desirable.
"When non-blocking, large optical cross-connect technology matures it might become a reality," says Diner. "MEMS technology is exciting and promising," he continues. "The only thing we don't have is wavelength conversion." In the few years between now and when the cross-connect technology becomes critical, Diner is hopeful that viable wavelength conversion techniques will be available.
By: Erik Kreifeldt