Cover head:All-optical, all the time

Having the bandwidth alone isn't sufficient. You have to manage what you've got.

And what you've got are plenty of cross-connects. Carriers rely heavily on their various digital versions. Digital cross-connect systems aggregate and groom traffic and send data on its way. They examine signals at the DS-3 (44.7 Mb/s) level and above without terminating it to the DS-0 (64 kb/s) level, making the network more efficient. They work with integrated or external add/drop multiplexers (ADMs) to hand off signals.

But there are drawbacks to the DCS approach. For example, the combination of network elements can be expensive. There is a greater chance for delay, and in some cases, failure. In addition, DCSs don't fit in a pure optical environment.

In an effort to circumvent those issues, the industry is looking to the optical cross-connect. It fits neatly into the future vision of the "all-optical network," even with its many definitions. But are optical cross-connects ready to revolutionize the network?

Maybe not today, but the time is coming. Optical cross-connects are evolving, even before they find a home in the network. But they are not easily defined-optical cross-connects fall into multiple categories. Some industry folks delineate devices based on switching matrix (electrical vs. optical), while others view functionality as the divider. Most optical cross-connects are still being conceived, but they hold promise and have the potential to change the way networks are built.

A cross-connect by any other name Scott Clavenna, an analyst with Pioneer Consulting, puts optical cross-connects in three classes, each category progressively adding management capabilities. Each type of device fits within a specific segment of the network. Although some refer to these optical cross-connects by different names, their functionality is essentially the same.

The most basic iteration of an optical cross-connect is an optical switch or fiber switch, Clavenna says. The fiber switch optical cross-connect operates in bulk fashion, routing all the optical channels from one fiber to another (Figure 1). It provides protection switching, restores signals without processing all the information on the signal and uses electrically based ADMs.

The fiber switch cross-connect does not handle individual wavelengths, or offer bandwidth management or traffic grooming capabilities. Fiber cross-connects are essentially optical protection switches. Available now, they are typically integrated into ADMs or wavelength division multiplexing (WDM) terminal multiplexers.

"It's a coarse level of protection in the optical domain," Clavenna says. "It's not a great value [add] because there is nothing it can do with an individual wavelength. The whole [link] has to go down before it recognizes [the wavelength is out]. There's not a lot of flexibility with it."

It might not be flexible, but it is useful, says Robert Pullen, vice president of engineering and marketing for Tellabs. Although fiber cross-connects have limited functionality, they are non-blocking. All incoming fibers or wavelengths are handed off as a single pipe to another fiber. This is useful in large networks to move vast amounts of data.

Indeed, Qwest Communications uses optical switches to hand off OC-48 (2.4 Gb/s) signals from its OC-192 (9.6 Gb/s) network to regional Bell operating companies and competitive local exchange carriers today, says Vab Goel, director of IP network engineering and advanced technology at Qwest. "It gives us simplicity in the network," Goel says.

The second type of optical cross-connect is a wavelength switching cross-connect (Figure 2). This switch goes beyond protection and grooms traffic to fill the pipe more efficiently. It can route a specific wavelength on one fiber to its corresponding slot on a second fiber. The catch? The right slot has to be available.

Switching individual wavelengths is a step in the right direction, but it has limitations, Clavenna notes. A finite number of wavelength channels exist, so a ring with many nodes might run out of available channels and block communications.

"If you put a wavelength on one part of the ring, that wavelength can't be used on any other part of the ring," he says. "It's a zero-sum game."

The true benefit of an optical cross-connect lies in its ability to manage bandwidth at the wavelength level. Enter the wavelength interchange or wavelength translating optical cross-connect (Figure 3). This device can take any wavelength entering the optical domain and convert it to a different wavelength for output. It is a non-blocking cross-connect, which Pullen believes will stimulate the world of true optical networking.

In fact, the wavelength interchange cross-connect might allow two signals to travel on the same channel in different parts of the network. "This is the ultimate cross-connect," Clavenna says. "The basic benefit of wavelength translation is that you increase the value of your network. In a multiwavelength network, you can manage wavelengths on each segment rather than just on entire networks."

That's an important factor for optical cross-connects, says Rod Alferness, chief technology officer of Lucent Technologies' optical networking group. Being able to cross-connect or drop optical channels changes the medium of exchange to wavelengths. And that means that devices can switch any type of wavelength regardless of the payload, he says.

All-optical confusion The term "all-optical" has a multitude of meanings. Vendors and carriers agree on the optical portion, but moving entirely away from electronics isn't always the most efficient method of moving or managing bandwidth.

"There is still some debate over whether or not you even want the optical cross-connect to be all-optical. That was the first panacea," Clavenna says. Originally vendors were striving to build ADMs, terminal multiplexers and switches that operated only in the optical domain, so the wavelength signal was never broken down.

Now, Clavenna says, "Vendors and carriers aren't quite as excited about the all-optical network as they were a few years ago. They are seeing value in having electrical processing occur at different points in the network. There are benefits to be able to examine the signal electrically."

One of the most important benefits is that an electrical matrix provides for Sonet reliability. Until something equivalent emerges in the optical domain, carriers will stick to Sonet, and vendors will implement optical-electrical-optical conversion.

Alcatel supports the vision of a photonic network rather than an all-optical network. "The big benefit of photonic switching is [that it operates] independent of rate and format," says Tim Krause, director of product line management for optical networks at Alcatel. But photonic switching can't regenerate signals or monitor performance. Today, regeneration requires electrical-optical-electrical conversion.

Krause predicts that all-optical switches will emerge in the metropolitan environment first. "In a metro application you don't need to worry about regeneration required in a long-distance network, but you do want a tremendous amount of transparency," he says.

Photonic optical cross-connects will move into the backbone later, adds Philippe Perrier, product manger for optical cross-connects in Alcatel's optical networks group. "It is difficult to make an electrical matrix work at 10 Gb/s, but an optical cross-connect with photonics can do it," he says.

Alcatel is testing prototypes of photonic optical cross-connects overseas. The company plans to ship product within the next two years, he says.

Other organizations have commenced work in that direction, as well.

For example, the Multiwavelength Optical Networking consortium is evaluating transmission, management, and control of wavelengths in the optical layer.

The Monet consortium includes AT&T, Bell Atlantic, Bellcore, BellSouth, Lucent, PacificTelesis and SBC Communications/Technology Resources Inc. and is partially funded by the Defense Advanced Research Projects Agency.

The companies constructed three geographically dispersed testbeds to evaluate optical cross-connects in local and long-distance environments. Wavelengths originating from an optical ADM on a local exchange network can be sent through an optical cross-connect onto a long-distance transmission path and back to the local environment. The Monet group also plans to connect to a new metropolitan WDM ring.

"It's a critical program in helping to define and demonstrate the feasibility of optical layer networking," says Lucent's Alferness.

A major hindrance to all-optical networks is that not all channels are equal. Vendor equipment often doesn't interoperate because each vendor can use a different part of the spectrum and implement different channel spacing.

Maintaining the electrical component preserves the openness of the network. Interoperability of vendor optical cross-connects is crucial to full-scale adoption.

"The standardization of this is peculiar," Clavenna says. "The technology is advancing so rapidly, the standards are just trying to catch up with this. The all-optical network is far enough away that most of the early standards of optical bandwidth management will probably be far along by the time we get to that."

Bellcore spinoff Tellium points out that the ability of its Aurora wavelength-converting cross-connect (Figure 4) to convert all incoming signals to 1310 nm for output allows otherwise non-interoperable systems to work together (see sidebar opposite). That saves money for merging networks and interconnection arrangements.

The perfect fit Companies are working to develop some type of standard technology for optical ADMs and optical cross-connects, says Larry Seese, executive vice president of network engineering at Qwest. Several issues still need to be hammered out, including how optical cross-connects will fit in the network, how they will be managed, who will operate them and what line rates they will support.

Before Qwest considers deploying a purely optical cross-connect, the device has to be commercially standard, Seese says. Other factors? "Advanced speed, bandwidth, simplicity, cost effectiveness," he says. "More important, do we know how to architect, engineer, implement and operate a network that is all-optical? Do we have the capability for performance, reliability, troubleshooting, managing [the network]?"

Bell Atlantic is pushing more optics into its network, says Mike Daigle, director of network planning, transport/circuit at Bell Atlantic. But to maintain its investment in its embedded base of DCSs, the carrier is adding optical interfaces in 1999.

Like Qwest, Bell Atlantic is largely concerned with the field readiness of the optical devices, as well as how optical cross-connects can be integrated with the legacy network equipment, operations support systems and element management systems.

"There is a wide variation in what you define as an optical cross-connect," Daigle says. "Some of [the optical cross-connects] are ready now, but they are optical interfaces on current technology. They do increase the functionality and usability of these machines, but it's not really a leap forward."

Daigle sees optical cross-connects fitting in the backplane, but not for a few years. The optical cross-connect will bring new functionality and new applications, he predicts. "We will have to rethink some of the design parameters as these devices become available. We don't have a specific answer as to [what] is the niche."

Seese also is optimistic about the future of optical networking. "An all-optical network gives you tremendous bandwidth applications," he says. "It will simplify networks, cost-reduce them. You'll also be able to work with huge chunks of bandwidth remotely," he says.

Today, Sonet is used for network performance and reliability, but if carriers can do optical switching and optical ADM functions without it, network costs might decrease by half, Seese suggests. In addition, optical cross-connects require less maintenance, fewer parts, fewer connections and less chance for error. Plus, bandwidth pipes can be optically assigned to customers, thus cutting provisioning costs.

When will optical cross-connects infiltrate the network? Not until they can ensure reliability and support high-speed line rates of OC-48 and above. Some devices can support OC-3 (155.5 Mb/s) and even OC-12 (622 Mb/s), but that isn't sufficient for the investment in optical cross-connects-especially as an emerging, non-standard technology.

"There is so much hype over optical cross-connects, we forget what is out there today. Similar [products] are doing similar functions, but you can make a more optimized product," Tellabs' Pullen says.

An optical cross-connect might sit on top of a traditional network or be deployed in a new network, but it's not going to revolutionize the way networks are built. At least not today.

"The optical cross-connect will be complementary to other items in the network," observes Mat Steinberg, director of optical networking at Ryan Hankin Kent.

As more functionality is incorporated, the economic factors come into play. Eventually it will make sense to deploy full-featured optical cross-connects in the network, but until an optical equivalent to Sonet emerges, carriers and vendors will rely on Sonet.

In the end, all-optical isn't pure optical.