technology & profits

What's in a name

Optical crossconnect? Optical switch? The terms may be confusing, but there's no confusion about the need for such equipment.

BY ROBERT PEASE

Many telecommunications carriers have been caught off guard by technology. First, there was the race to get more and more bandwidth as data and the Internet sent capacity demands soaring. The fix to the problem was the advent of dense wavelength-division multiplexing (DWDM), enabling carriers to increase the capacity of a single optical fiber to enormous levels. But with all those wavelengths transporting traffic to all the required locations, carriers are faced with a bigger problem-how to manage them.

The need to manage fiber-optic networks capable of near-limitless capacity is creating a photonic switching market that will average a 45%-per-year growth rate through 2003, according to a report from ElectroniCast Corp. The San Mateo-based telecommunications consultancy expects the market to reach $761 million by 2003 and continue to $4.11 billion by 2008.

"The growth will be driven by the need to implement switch functions that do not limit the capabilities of the optical fiber," says Stephen Montgomery, president of ElectroniCast. "Current OEO [optical-to-electrical-to-optical] switch limitations include limited data rate, sensitivity to data protocol, complexity and cost for multiple wavelengths, high power consumption, and the inability to transparently pass all of the optical signal on the fiber."

Additionally, Montgomery and the firm's chairman, Jeff Montgomery, say improvements in technology, such as the maturation of optical micro-electromechanical systems (MEMS)-based solid-state switch components, will provide higher performance, higher reliability, smaller size, and photonic matrix devices with low power consumption at a significantly lower cost per channel.

Simply stated, optical crossconnects or switches (depending on your preference of terminology) are designed to terminate incoming optical signals and perform crossconnections among any number of ports on the device through a switching matrix, either electrical or optical. The optical-switching market mainly comprises optical switches, optical crossconnects, and almost any hybrid of crossconnect, switch, router, or other box that basically performs the switching function. Deciding what terminology best describes the product is a marketing issue. Determining whether today's OEO switches or tomorrow's all-optical devices will adequately meet your needs is perhaps more important-but probably no less difficult.

Pioneer Consulting forecasts steady growth in optical networks through 2004, significantly impacting the long-haul and metropolitan markets for optical-switching equipment. As the bandwidth increases in these optical networks, carriers are experiencing an immediate need for switching solutions that can effectively manage.

Until recently, finding the right piece of equipment for an individual application was relatively straightforward. If you needed a crossconnect function performed, you would buy a "digital crossconnect." If you needed routing, you'd buy a "router," and if you wanted switching, you'd buy a "switch." Today, however, finding agreement on the definitions of such terms as "optical crossconnect" or "optical switch" can be difficult at best, since these terms are often used interchangeably. Take for instance, "optical crossconnect." Dana Cooperson, a senior analyst with RHK, says her firm tried to define switch fabrics that switch at STS-1 (52 Mbits/sec) or below as digital crossconnects. Those that switch wavelengths at OC-48 (2.5 Gbits/sec) and higher would be referred to as optical crossconnects, regardless of whether the switch fabric was optical or electrical.

"Now, we've got several companies that don't meet just one criteria or another, so it's a question of how finely you want to cut this thing," says Cooperson. "With each box having a different level of restoration, is it really a switch or a crossconnect? There are various shades of gray along that dimension as well. Right now, when someone says they've got one, we ask them what they actually have, how it really works, and determine what box would we put it in."

Cooperson says that sometimes the label a product carries may serve merely as part of a marketing initiative, such as putting the word "Internet" in front of something. By putting the word "optical" in front of a name, the hope is that people will be more interested and excited about it.

"You really have to scrub out the hype and get to what they really have today and what they promise to have in the future," adds Cooperson. "But no matter how you label it, we see it as one of the three big growth areas in optical networking."

A similar problem exists in differentiating an optical crossconnect from an intelligent optical switch. "Intelligent optical switch" is a term used loosely by several equipment vendors. ElectroniCast defines "optical switch" as a device that performs switching among optical input and output ports. An "intelligent optical switch" is defined as a network element that couples the functionality of an optical-switch device with some or all of the features of Synchronous Optical Network (SONET) add/drop multiplexers (ADMs), broadband digital crossconnects, and data switches.

"Though these sound similar to the hybrid DCS (digital-crossconnect system)/ADM offered by major SONET vendors, they differ in their scale and their focus on switching high-bandwidth circuits," says Stephen Montgomery. "The intelligent optical switch must support at least 32x32 switching and provide a migration path to 1,024x1,024 switching for the near future."

Scott Clavenna, an analyst with Pioneer Consulting (Cambridge, MA), agrees that in the end, an optical crossconnect can groom to a higher level than a DCS. The fact that optical crossconnects can switch at optical line rates, normally above OC-3 (155 Mbits/sec), means the element lives up to the optical name.

"For now, though, most seem to prefer the term 'optical switch' as opposed to 'optical crossconnect,'" says Clavenna, "but they're almost interchangeable terms in the industry now."

Clavenna and others add that two camps have formed in terms of optical switches-opaque and transparent. The opaque optical switch operates at OC-48 and OC-192 (10-Gbit/sec) line rates, but the functionality is performed electrically; thus, data must be converted out of the optical domain for at least part of the process. The transparent optical switch (or "all-optical switch," in the parlance of some), ideally requires no electrical conversion.

In the opaque camp are the "big five" in terms of optical-switch vendors-Ciena Corp., Sycamore Networks, Tellium, Cisco Systems, and Nortel Networks. These products are based on the electrical conversion technology that is carrier grade and now commercially available. On the other hand, startup visionaries like Corvis, Qtera, and Chorum, and others that have promoted all-optical-networking architectures largely drive the transparent camp.

"Although they haven't announced very detailed information on the optical-switching platforms they have, they do seem to be transparent," says Clavenna. "They don't have the same density as the opaque switches. Instead of 256x256, they're only 8x8 or 16x16, but they do perform that optically. Fujitsu and Lucent Technologies are also in this camp, with Lucent's MEMS-based LambdaRouter leading the way with equivalent density with the opaque switches. Although this product isn't ready for prime time yet, it does show where this camp is headed."

The analysts agree that, in theory at least, the upcoming generation of all-optical-switching equipment will greatly improve on today's OEO technology, particularly as the achievable matrix sizes grow and carriers and vendors can iron out the details of how a truly all-optical network might work. This factor poses a "when to deploy" problem for carriers in a competitive world where timing is everything. The fundamental problem of managing an ever-growing number of wavelengths across fiber-optic networks has carriers caught in the middle. Do they deploy what's available now, or can they possibly postpone major architectural changes in order to wait for the promised advantages of the all-optical, transparent technologies?

"I think it's always a struggle for the carrier to decide whether it's better to wait to get something better versus having an immediate problem that needs to be solved now by any means that makes sense," says RHK's Cooperson. "The pain from not solving the problem may be worse than getting something in the interim that may not be state-of-the-art in a year or two. I think there's always that 'when do I buy' thing going on."

For the immediate "pain" the long-haul carriers are experiencing, the opaque optical switches offer at least a viable remedy, if not a cure. All of the major vendors in this camp have products with a growth path to 1,024x1,024. Used at OC-192 rates, these products offer a massive switching system that many carriers feel would tide them over for at least three to five years.

"These are really not just interim products," says Clavenna. "They are a 'quick fix' kind of product. Carriers have been waiting for these products to fix their network congestion problems and inefficiencies. It isn't a technology that's looking for a solution; it's a real demand in the network. There's an immediate, pressing demand to start switching at line speeds."

Even if the equipment is ready, there may still be questions revolving around whether the accompanying "smart" software is ready in terms of restoration and provisioning. According to research at Pioneer, the switches supplied by Cisco, Ciena, Nortel, Tellium, and Sycamore differ in their support of grooming lower-bandwidth circuits and to what degree they can switch signals below OC-48. All should be capable of providing a variety of restoration algorithms, including 2- and 4-fiber bidirectional line-switched ring, unprotected, or mesh-based restoration.

The switches will be expected to support 50-msec restoration algorithms when the service requires it and cost-effectively deliver bandwidth to applications that manage their own protection schemes at other layers of the protocol.

The ability to offer flexibility in restoration schemes allows carriers to more efficiently tie together their overlay voice and data networks or build more elegant network architectures from the ground up. Capabilities such as these are what many vendors refer to when they discuss "intelligent" optical switches.


But when push comes to shove in terms of necessity in managing more capacity, carriers may be forced to take immediate steps with available products to alleviate bottlenecks despite the promise of what's to come.

"We think the problem for carriers is acute enough that they're willing to take a 'dumber' solution sooner if they can get it and if the vendor has a good capability for expansion, both in port number and software-related restoration capabilities," says Cooperson. "If the roadmap looks good and they feel comfortable with the vendor, they're likely to try and go sooner rather than later. There's enough skepticism on the part of the carriers about the mesh-restoration mechanism that I don't think they'll let that stop them from deploying. The fact of the matter is they need something to get them out of the hole they're in now, so I think they're ready to jump-maybe not with both feet, but they're willing to try it out."

What will the carriers miss by not waiting for the all-optical solutions? The advantages of a pure optical-switching mesh include greater flexibility in their support of any variety of optical signals, higher density and, ultimately, higher switching capacity than electrical switch cores. The all-optical-switch fabric will provide higher scalability, where today's electrical switch fabrics have to grow in stages and will come with very big footprints once past the 1,024x1,024 matrix size.

"New carriers are probably more likely to adopt the transparent optical switches into their network architectures of all-IP [Internet protocol]," says Clavenna. "They'll have terabit routers really calling the shots in terms of intelligent-service aggregation and management."

There may also be opportunities for co-existence of opaque and transparent optical switches within the same architecture in large networks, according to Clavenna. However, while Cooperson sees the two possibly working together, she points out that dislodging fundamental elements of a network, once in place, is difficult. That may be particularly true of the OEO crossconnects.

"Once you have the embedded base and initial investment, it's comparatively easier to add a bunch of ports than to scrap it and move to something else without a compelling reason," says Cooperson. "We see this as a race. We also see them [opaque vs. transparent] as competing technologies, not so much as complementary technologies. If the OEO switches get in this year or the beginning of next year and gain momentum, I think the carriers will find it harder to switch to something else without some really compelling reasons to do so."

The pressure is on the all-optical transparent camp to make products available before carriers make their move to solve their network bottleneck problems. However, the research by Pioneer, ElectroniCast, and RHK suggests that "prime-time" products are still not commercially available, although most vendors in that camp are still rather tight-lipped about their all-optical product portfolios.

ElectroniCast suggests that "it will not be possible for the system builders to continue their wait for the promised perfect solutions. The maturation of photonic-switch technology is likely to take more than 20 years. But in the short term-two to five years-the first products that are able to meet the minimum requirements will have to be adopted."

Meanwhile, market opportunities exist for the opaque optical switches that, according to Pioneer, have broader applications then originally anticipated. Clavenna suggests that DWDM networks are not necessarily a requirement for optical switches.

"You could really see ILECs [incumbent local-exchange carriers] and CLECs [competitive local-exchange carriers] using these in certain metropolitan areas," says Clavenna. "Wherever you have stacked SONET rings terminating at one hub, these systems can perform some very efficient switching. It doesn't necessarily require DWDM in that sense, so it's not just an outgrowth of the DWDM market. It's really an answer to a lot of problems resulting from the growth of SONET fiber networks in general. So I think there is a tremendous market emerging there that really hasn't been quantified and extends much further beyond just the core switching of IXCs [interexchange carriers]."

On the heels of bandwidth expansion, wavelength management has become a critical issue for telecommunications carriers. Vendors came up with lots of solutions for creating enormous amounts of capacity and now the challenge is to manage it. Whether you call them optical switches, optical crossconnects, or anything in between, these network elements have their work cut out for them and will play a major role in intelligently managing fiber-optic networks- today and tomorrow.