Improving the economics of a network build was the motivation for proposing an OO approach. But carriers are facing a "CapEx conundrum".

By Jennifer Sorosiak, FPN Contributing Editor

When it comes to all-optical-to-optical (OO) networks, one vision is evidently emerging: Aggregate services at the edge, carry that traffic "in optics as far as you can to get the best price performance, and ... disaggregate for service management or service injection back further in the network." Peter Evans, senior vice president of marketing at ONI Systems (San Jose, CA), suggests that while an OO network "has architectural benefits, the technology hasn't quite progressed to provide the level of network cost and performance management capabilities that you can readily get from an optical-to-electrical-to-optical (OEO) application, or has yet to fully develop the transition plan to allow carriers to leverage their existing networks." Perhaps this is why carriers are facing a "CapEx conundrum".

"The prevalent thinking is to deploy the best technology that addresses current service needs. OOO systems, while compelling, (don't currently provide for) the price points and management technologies where systems are mass deployable and become competitive with substitute (OEO and OEOOEO) products that already exist in the industry."

Clearly, a conundrum is where this story begins.

Promise
In pure optics, explains Richard Norman, CTO at Hyperchip (Montreal, Quebec), "things stay photons the whole way through.... If you have a pure optical core it means at the edge of the core you'd have electronics and you would never convert from photons to electrons again until you reach the far edge of the core. You would have amplifiers, but no regenerators, and that would mean that you'd have OOO switches if you did any switching...."

Transparency, meaning "freedom from headaches," is the benefit of an all-optical approach. When coupled with scalability, Norman says, a low future cost is "the practical benefit".

An OO consists of a photonic switching fabric, using one of a variety of switching technologies.

OEO switching incorporates an electronic switching fabric with photonic interfaces.(diagrams courtesy of CIENA)

Improving the economics of building a network was the motivation for proposing an OO approach. Tom Mock, vice pesident of Product Portfolio Management at CIENA (Linthicum, MD), which plans to acquire ONI, explains that unfortunately when the vendor community "started looking at (OO) we didn't figure in the costs. If I have two optical routes that I can switch between, those routes may require, for example, different dispersion compensation. If I predicate that that network is going to be dynamic, then ... dispersion compensation — and other types of transmission impairment compensation — also need to be dynamic.... There are certainly ways of doing that, but with today's technology, they are not necessarily cheap. Overcoming those issues (without complicating deployment) is one of the key elements to making an all-optical network economically viable."

"There still is a high level of interest in moving towards an all-optical network as the natural convergence layer." According to Evans, all-optical "addresses many existing and emerging service needs and can carry these services in ... their native format".

When it comes to next-gen optics, cheap and cutting edge are not always mutually exclusive goals. "Maybe on a port by port basis," Evans explains, a "next generation optical system is a bit more expensive. But, when you architecturally look at how you aggregate and manage the whole varied set of services across the network, next generation optical networks are actually a more compelling way to get momentum out of your CapEx dollars."

Regarding traffic on an inter-office facility, Evans notes, "at about two OC-48s worth of traffic one reaches the point where it's more cost effective from an operating expense (OpEx)" as well as from a CapEx view, "to start deploying a next-gen optics versus a traditional SONET network."

"People want to maximize the use of their existing equipment. They don't want to bring in a forklift. They want to ... add new equipment that enhances the way their old equipment works," Norman adds.

The current "pause in the industry has been a good thing because it has allowed people the opportunity to really think about how they want to evolve their network. Previously," Evans suggests, "people thought about revolution and laying new networks on top of existing ones." Evolution is now the prevailing view.

In the past there was "an atmosphere of 'let's put in the latest and greatest technology, damn the torpedoes, full speed ahead,'" says Mock. "That kind of environment certainly doesn't exist any more."

Maximize
When considering a network build, Mock argues "there are some real economic advantages to be had from grooming that aren't necessarily intuitively obvious. If you can better fill the pipes in your network, for example, then you can build fewer pipes. Often times the cost of the pipes is larger than the cost of the switching element..."

Function consolidation, Mock explains, has proved crucial to improving the CapEx side of the equation. Increasingly, carriers and vendors are focusing on OpEx reductions. There "is a need for these different provisioning devices, be they all-optical switches or grooming switches, to be able to communicate with each other, to be able to establish connections through the network automatically. ... It's something that we see as key to making the economics work moving forward."

Even with a CapEx slowdown, bandwidth demand suggests "there will be about a ten-fold increase in the number of OC-48 'equivalents' of traffic." Evans says, "Whether it is voice, data or other services, over the next five years ... you start to realize that the economics of continuing to deploy platforms that are based on the economics of heritage technology really doesn't prove out in the long term."

Evans provides an historical perspective supporting his claim. "About 10 years ago, 2.5 gigabit SONET systems were prevalent. The average system consumed about 280 watts per gigabit of capacity and you could get about a gig of capacity in a square foot." Around 1995, 10 gigabit systems were introduced. "They were probably drawing about 100 watts of power per gig," he says, "and getting somewhere in the range of 75 gig of capacity within that square foot. (Today) we've pushed the envelope as far as we can with those SONET boxes and (are moving to) next generation optical platforms." Such solutions, Evans explains, draw "less than about 10 watts per gigabit and are in the range of 200 gigabits of capacity in a square foot. Based on pure physics, power, and real-estate consumption ... a carrier can address new service needs and drive further OpEx reductions."

Today, "people are going with what feels comfortable and will continue to buy some legacy systems." Evans contends, "They are looking for and deploying specific applications of next-gen optical systems. They are (also) looking to find ways to make sure that (such) systems are integrated into existing heritage networks..."

Optical Switching
One should use optics, Norman says, "for what they are good at, and use electronics for what they are good at. Electronics are good at high touch per bit. They cost a lot but you can do a lot on every single bit. Optics simply can't compete on high touch, but the goal is to do the electronics as few times as possible."

Norman suggests, "Instead of the 17 to 18 router hops that you currently end up with on the average because today's electronics simply don't scale," next-gen technologies enable one to get hops "down to four. The way it is done today is seriously ugly; the average packet is routed 10 times within the core alone."

"The ability to do as much aggregation as you can at the edge of the network and (to) wrap the new services and protocols in SONET ... provides a simple evolution path (to) feed into your existing SONET (service management) network and (to) leverage the existing investment." Evans adds, "Being able to then augment with all-optical or wavelength systems to not only carry those services in those places where you don't have the SONET bandwidth, but to then carry services natively also provides further opportunity for growth."

While Mock says relatively expensive "optical switches themselves are ... limited in size (and) somewhat unproven when it comes to reliability," he believes they have a place in a network. "Probably there will be two somewhat different roles: One in the core of the network and one toward the edge.

"Core optical switches," continues Mock, "are likely to be larger. They may start at something relatively simple and then evolve into what we call a multi degree switch that has traffic coming in from multiple directions. In the local network, just because of the way that the fiber routes are architected, I think you'll see generally smaller switching devices with lower price points and lower emphasis on lower first installed costs..."

A few companies, Norman says, are making all-optical switches. "The Corvis optical switch is truly optical at the heart. A Lucent Lambda router is optical at the heart ... but you need to electronically re-boost the signal just before you go in. A Sycamore switch is definitely OEO. Companies that are doing the MEMs (micro electro-mechanical) work are coming along." He says the questions for none-purists remain, "Who cares whether it's really OEO or OOO? Who cares what color a cat is as long as it catches mice?"

Fundamentally, "end user demand really hasn't changed. In a lot of cases people have," according to Mock, "either (over) built capacity or they have built networks that weren't necessarily very efficient. Now, more attention (is) being paid to making sure that economic models work as well as they possibly can. There is a fair amount of attention being paid to supporting new services, like Ethernet, but those goals aren't necessarily incompatible with an OEO type switching fabric." He adds, "To the extent that an all-optical network might force me to put a one gigabit signal on a wavelength rather than a 10 gigabit signal, I can get better use, particularly out of my long distance network, if I aggregate traffic up before I put it on the base transmission network."

Mock contends, "People have also recognized that with the proper software, grooming switches (OO and otherwise) can actually do more than just switch traffic."

In the Midst
"Next generation," Norman notes, "is basically saying (that) in the long haul you keep it photons all the way across the continent ... rather than going through all the older gear that needs to incorporate re-gens every, say, 300 kilometers. Because the distances are short enough that one doesn't need a re-gen at all, metro transport will get to the all-optical nirvana before the long distance does. Only where the (access and long-haul) domains meet is where you should have electronics."

Improved economics was the promise that initially spurred interest in the pursuit of all-optical networking. Enthusiasm will again be expressed "when it begins to make economic sense," Mock says. "Unless that really happens, and unless the technology is really proven to be carrier class and reliable, carriers will still be a little shy of it."

The reality is, continues Mock, "there are really very few applications out there that require a completely all-optical network.... There are some advantages to doing that in terms of transparency. In many cases those are outweighed by the cost of doing it with today's technology. A grooming switched network and an all-optical switched network will in fact co-exist because there are places in the network for both types of technology. The key to that is going to be putting in place a network-wide control plane that allows them all to interoperate properly."

"It's going to be a while before people do pure optics for the sake of pure optics. They will do pure optics for the sake of OpEx, not optics," Norman notes. "(With) pure optics, once they are in place you just let them run. They don't care what they are transporting. They are ultra reliable because they have no moving parts. The carrier doesn't care if it is pure optical or OEO. They just want to get their job done with the minimum of headaches and at minimum cost."

---

Cahners, FPN,