Pragmatic look at optical-switch market

BY MEGHAN FULLER

While the advent of DWDM has drastically increased the number of channels on a single optical fiber, a simple increase in raw capacity does not give service providers and carriers enough of a competitive edge; they need some sort of traffic cop to route the channels to their diverse destinations. Today, carriers have a lot of DWDM point-to-point long-haul transport systems in place, and at the junction sites where SONET rings meet, they need to deploy some kind of intelligent device capable of handling the capacity.

"At every place where those rings connect, they hit a digital crossconnect or legacy SONET equipment, and a work order has to be sent out; a network operator has to be sent out and physically make that crossconnect at every switch along the way," explains Karl Traberg, director of marketing for the terrestrial networks division of Alcatel (Paris). This process can take six to nine months. Moreover, the traditional SONET rings allowed for simple point-to-point connections, but to really get the most out of DWDM, a mesh architecture should be implemented.

Enter the optical switch, a device that enables faster provisioning, faster restoration, and the ability to migrate a traditional ring architecture to a mesh architecture. The photonic- or optical-switch market could reach an estimated $4.11 billion by 2008, claims telecommunications consultant ElectroniCast (San Mateo, CA).

Today's optical-switch market is certainly dynamic. Analysts, vendors, and even the end users themselves have made wild and wooly predictions. As a result, it's difficult to determine where the marketing hype ends and the actual product begins.

Before we can even begin to separate fact from fiction, however, we need to get the terminology straightened out. Many in the industry use the terms "optical switch" and "optical crossconnect" interchangeably. Is the difference just a matter of semantics, or are there physical or functional differences, as well?

According to Krishna Bala, president and CEO of Tellium Inc. (Oceanport, NJ), optical crossconnects and optical switches are one and the same. "When you say optical crossconnects, people will associate that immediately with digital crossconnects, and they start saying, 'Oh, okay, it's a dumb patch panel,' things like that," he says. "We wanted to get away from that, so we invented this new terminology, but the truth is today, there is absolutely no difference."

Stephen Ferguson, director of photonic market strategy at Marconi Communications plc (Pittsburgh) disagrees-to some extent. While the basic engineering of the two might be similar, he defines a crossconnect as a device typically controlled by the operator, while a switch is typically controlled by the end user or the application. John Adler, director of marketing, wavelength-routing business unit at Cisco Systems (San Jose, CA), takes that definition one step further, arguing that an optical switch is a device that switches wavelengths that are gigabits or higher, while broadband crossconnects switch at the megabit level.

According to Rick Thompson, director of product marketing at Sycamore Networks (Chelmsford, MA), "a crossconnect takes the interfaces and makes them higher speed, takes the footprint and makes it smaller, reduces costs-all of which needs to occur," he says. "A switch does all that plus integrates intelligence in the software. So the real difference between the crossconnect and the switch is the software level of intelligence the product has."

No one disputes the need for intelligence in such photonic devices, whether they're called crossconnects or switches. "Crossconnect or switch, the real question to ask is, 'Do you provide dynamic point-and-click provisioning without manual intervention?'" says Bala. "If the answer is yes, then you have advanced software capability, and [the issue of] switch or crossconnect is moot."

At its most basic level, an optical switch or crossconnect is used to add and drop channels and switch them between incoming and outgoing ports. Its switching fabric may be either electrical or optical.

Optical-electrical-optical (OEO) switches compose roughly 95% of all switches currently deployed, says Bill Wier, vice president of marketing at Power X Networks (San Jose). These switches convert the optical signal to electrical for performance monitoring, grooming, and what Traberg calls "3R functionality": retiming, reshaping, and reamplification. When a signal exits an optical switch, it may be differently structured because traffic has been switched around, but the cleanliness or purity of the digital signal has not changed.

Today, all of the intelligence resides in the electrical world. Once a signal is converted from photons to electrons, "you can go in [to the payload] and tear apart the packet, look at its destination, look at where it came from, what kind of packet it is," explains Wier. "Is it voice? Is it data? Is it video? If it's all those, which takes priority? And how much of it is there? Should there be reservations for such priorities?" Because of this intelligence, OEO switches allow point-and-click provisioning in seconds rather than months.

If OEO switches are so great, why even bother with all-optical, or OOO, switches? First, the optical-electrical-optical conversion adds expense to the network and can create a bottleneck in transmission. OEO switches are content-dependent. "If you are converting something into an electrical signal, you need to know what bit rate it is, and in some cases, you have to know whether the signal is synchronized, whether it is framed, etc.," explains Demetri Elias, vice president of marketing at Sorrento Networks Corp. (San Diego).

Second, OEO switches just aren't as scalable as OOO switches promise to be. Given the increase in Internet Protocol-centric traffic, "a typical long-haul carrier in the U.S. just a few years from now will need something beyond 50- or 100-Tbit switching capacity in their core nodes," says Thompson. He believes the technological limit of electrical switching is 1,000 ports of 10 Gbits-or 10 Tbits.

All-optical switches, on the other hand, have the potential to scale much higher. "We see 512x512 ports coming this year," he says, "and 1,024x1,024 by next year, which is 10-Tbit capacity. From that point, it will grow further. It can grow beyond 100 to 200 Tbits with an all-optical switch fabric."

Applicable at the highest levels of bandwidth, in central offices where fat pipes are coming in and out, all-optical switches are transparent. "When you do all-optical switching," explains Elias, "you are not constrained by bit rate, you are not constrained by protocols, you are not constrained by any optical-to-electrical conversion."

All-optical switches are also incredibly compact for handling high bandwidth in terms of scaling, power consumption, and rack footprint. "I did some comparisons of technologies being commercially offered, and I found that there was of the order of five to 10 times difference between an OOO and an OEO nominally doing the same switching function," says Ferguson.

OOO switches have more than their fair share of operational and engineering obstacles, however. For starters, there's no way to convert light from one wavelength to another all photonically. "Picture in your mind two wavelengths coming into St. Louis from the West Coast," explains Joe Bass, vice president and general manager of wavelength routing at Cisco. "They're on the blue wavelength, but we only have one blue wavelength available from St. Louis to Chicago, our destination. Which one of those do we not take through the network?" To circumvent this problem, network operators would have to engineer wavelengths through their network knowing or anticipating which colors they are going to need from which direction, hardly a practical solution. Today, wavelength conversion can only be done in an electrical format.

In addition, all photonic switches do not have a built-in regenerative function. What you have with an OOO, explains Marconi's Ferguson, is essentially an analog machine; a signal goes in clean but does not come out that way, due to a buildup of noise distortion and other errors.

Operational challenges aside, there are still basic hardware challenges that have yet to be overcome. There is no consensus among vendors about the best optical-switching technology. The search for a technology that provides the best economics and scalability has become the industry's hottest trend. Lithium niobate, thermo-optics, semiconductor optical amplifiers, and thin-film or polymer devices have all, at one point, risen to the forefront of this debate. Today, however, the three most promising options seem to be liquid-crystal technology, Agilent's inkjet bubble technology, and micro-electromechanical systems (MEMS).

The strength of liquid-crystal technology rests in its proven reliability; it has already been embedded in other more complex devices, including optical add/drop multiplexers. Agilent's inkjet bubble device, the technology of choice at Alcatel, is relatively inexpensive and easier to manufacture with a high yield. Neither of these technologies provides the scalability of MEMS, however.

MEMS has emerged as the early frontrunner; market researcher CIR Inc. (Charlottesville, VA) expects the market segment to be worth $2 billion by 2004. Tellium's Bala admits seeing "a glimmer of hope in MEMS," but warns that even this technology has "three big hurdles that mankind has to transcend before we can declare a victory for 3D MEMS." Reliability and control are the biggest hurdles, followed by packaging and manufacturing. "These feature a micro-array of mirrors aligned to micro-lenses aligned to micro-arrays of fibers," says Bala. "It's not that straightforward."

Even if the hardware challenges are overcome, the software challenge remains. Before OOO switching is wholeheartedly embraced, vendors have to find a way to incorporate the intelligence of the OEO switch with the scalability of the OOO switch.

"We don't yet have a religion on technology, whether it's OEO or all-photonic," admits Bass. "What we're trying to do is solve real world practical problems in a manner that's cost-effective and repeatable on a large scale."