Discerning the 'all' from the 'almost' in all-optical networks

By ROBERT PEASE

TECHNOLOGY

Optical equipment manufacturers, new and old, are aggressively pursuing technology that will remove electronics from fiber-optic networks in favor of all-optical gear to make the network less costly and more transparent. Yet, it is arguable that there isn't now, nor will there ever be, a complete end-to-end all-optical infrastructure in which optical-electrical-optical (O-E-O) conversions won't be needed. So is the "all-optical network" a lot of hype or simply a matter of definition? According to three manufacturers of optical switching gear, it's probably both.

Today, electronics are a fact of life in an end-to-end network. Simply put, there is currently no means to discern if a particular information bit is a 1 or a 0 without electrical conversion. Thus, communication signals transmitted over fiber-optic wavelengths generally require O-E-O conversion at some point across the network, particularly if regeneration of the signal is required. At a recent trade show, representatives from Ciena Corp. (Linthicum, MD), Calient Networks (San Jose, CA), and newcomer Luxcore (Atlanta) formed a panel to address whether this requirement can be changed.

Part of the drive toward an all-optical network is getting the jargon straight. "What we mean by 'all-optical network' is a network in which the signal you launch stays optical the entire time it's in the network," says Steve Alexander, senior vice president and chief technology officer at Ciena. "This prohibits the use of O-E-O conversions within the network. This is the strict definition of an all-optical network."

The achievement of the all-optical network by this strict definition requires taking each function of the traditional network and finding the means to perform that function optically. Whether the recent introduction of switches and switching fabrics targeted at the all-optical network represents a huge stride forward, again depends on your definition of terms, the speakers agree.

"When I speak of a photonic switch, I think it's a unique term," says Charles Corbalis, president and CEO at Calient, "and I'm not sure we've actually seen it yet in our generation. We normally think of switches as packet switches or frame relay switches or ATM switches or SONET/TDM switches. We don't think of something that actually switches. But that's what a photonic switch is. It basically switches light and allows you to interconnect these fibers, one to another. Provided the photonic switch is as transparent as the fiber connected to it, it should be an incredibly high-value proposition for the service provider. It means you can put it in once and keep it there forever with no reason to ever pull it out and replace it in the future. You may need to change the interfaces for different applications, but the photonic core can remain the same."

Calient recently announced a 256x256 all-photonic switch based on micro-electromechanical systems (MEMS) technology. Corbalis points out photonic switching systems are basically lower cost and substantially lower power than their electronic counterparts, adding other cost benefits on top of the basic cost savings of equipment.

"The reason is clear," says Corbalis. "If a photonic switch is transparent across the same range as the fiber, it really doesn't cost any more to switch an OC-3 [155 Mbits/sec] or an OC-768 [40 Gbits/sec]. It's the same cost to switch one wavelength or a bundle of wavelengths, and nothing has to be changed to move a customer from a Fibre Channel interface to a Gigabit Ethernet interface."

As photonic switching moves from the laboratories to the service providers, it should also break the overbuild syndrome that service providers face. All-photonic switching that offers an attractive "one size fits all," futureproofing dimension to carriers' infrastructures could go a long way in making all-optical networks a reality.

Despite Corbalis's assertions, many in the industry would state that all-optical networking systems, to some extent, have been tested, deployed, and used for years, at least in one form or another. Running traffic over fiber for long distances without electronic regeneration is, for some, a loose definition of the all-optical-network architecture. Companies like Corvis Corp. (Columbia, MD), Qtera Corp. (Boca Raton, FL), and others provide all-optical gear that carries traffic across long distances without electronics, keeping the traffic in a purely optical state from end to end. Corvis recently announced a 4,000-km all-optical transmission over Broadwing Communica tions' (Cincinnati) fiber-optic backbone.

"People have gone from Europe to the U.S. all-optically for some time now," says Alexander. "It's called a submarine cable. Another area that has received a lot of attention and may have a lot of economic benefit, particularly in access, is the passive optical network [PON]. You can make a good argument that a PON is a version of an all-optical network. Essentially, you just have electronics at the edge of it and, internally, it's entirely optical."

Alexander says work has been done in North America, Europe, and Asia for years to eliminate electronics from as much of the network as possible. The focus in past years was primarily on eliminating regeneration equipment from the long-haul segments of the network. Corvis mesmerized SUPERCOMM 1999 attendees by announcing a successful optical transmission over a 3,200-km distance without electronic regeneration-thus, according to Corvis, the all-optical long-haul network was a reality.

But the years leading to the Corvis announcement, says Alexander, taught all-optical proponents some lessons. "Distance wasn't the issue, it was connectivity," says Alexander. "AONs [all-optical networks] just don't scale very well. If everybody gets a wavelength, you'll quickly run out. Today's state-of-the-art networks generate about 100 wavelengths, but even if you get to 1,000, it still isn't enough. You need millions of connections."

Today, the only way to get such networks to scale is to deploy wavelength-selective switches and wavelength converters. The network needs the ability for traffic to enter the switch on one color and exit on another. The only way to build a network that scales is to have O-E-O conversion somewhere in the network.

"In general, the AONs that can be built today aren't really what we want," says Alexander. "What we live in today is an 'analog' all-optical world. The fact that information bits cannot be determined optically means that there are electronics somewhere in the network. For me to tell you whether you have a '1' or a '0' present, I have to bring it down to electronics."

A "digital" all-optical network would require the ability to determine the bits on the fiber using optical interactions-and we're not there yet. Some of the missing pieces include optical memory, optical packet header recognition, and true photonic switching on a photon level.

But even given the limiting factors for all-optical networking in its purest sense, the industry still ventures farther and farther down that all-optical path. Long-haul segments can be done all-optically. Add/drop multiplexing can be performed all-optically. Local and metropolitan networks are being built all-optically, where PON networks provide "islands of transparency." Network architects have even worked out optical protection switching.

Alexander envisions a "three-hop" network architecture. One all-optical hop is from the point of origin to a switching site. A second all-optical hop travels along the core network to any location on the backbone. The final all-optical hop is used to move from the core to the final destination. The entire transmission is completed with just two O-E-O conversions, one at each edge, enabled by a hybrid O-E-O/optical-optical-optical (O-O-O) switch.

"The principles behind AONs are really simple," says Alexander. "What they say is: If you don't have to look at the bits, stay optical as long as you can and go as far as you can. That, generally speaking, is the cheapest way to build a network. But if you need to touch the bits anywhere in that network, you'll need electronics to do it."

How do we go from a hybrid all-optical network to a pure, all-optical, end-to-end network? Startups like Luxcore feel that disruptive, innovative technologies could cross the boundaries that now limit optical transport.

"The difficulty lies in being able to manipulate the control information within the switch fabric," says Gerald Ramdeen, CEO at Luxcore. "We still need to instruct devices throughout the network to open and set up paths. So we believe the first generation of these systems will have some electrical regeneration requirements, but they should be minimal and strictly for processing control information."

However, Ramdeen still believes that all-optical networking-in its purest, non-electronic form-is inevitable. "There are techniques being developed and worked on in the research environment where this control information could be kept all-optical," says Ramdeen. "There is also the development of systems and components that will be able to process that information in its optical realm."

Carriers need to look toward new techniques for all-photonic routing and switching, ultra-long-reach amplification, higher channel counts and throughput speeds, and dealing with issues of dispersion and noise. There are benefits to all-optical technology that go beyond simple dollars and cents.

"The biggest benefit, in our opinion, is transparency," says Ramdeen. "Carriers are getting tired of having to replace gear every 24 months to adjust to higher and higher speeds. OC-48 [2.4 Gbits/sec] and OC-192 [10 Gbits/sec] is prevalent today, but you have DWDM systems on the bench and folks that are doing work at OC-768 [40 Gbits/sec]. Carriers don't want to endure forklift upgrades or the massive replacement of electronics to support those higher throughput speeds. If your transmission and control data is all within the optical realm, supporting higher and higher bit rates becomes an easier task."

The consensus is that a lot more work lies ahead in achieving the true all-optical infrastructure. "It's still not clear that all-optical networking will end up being cheaper," says Alexander. "The point is that we still have a lot of work to do in building truly digital optical networks-and, meanwhile, we continue to fight the analog nature of them."

More in PON/Systems