The great light hope

All-optical networks are more than just a pipe dream, 
but the transformation won't be smooth

The XFL. The "Planet of the Apes" remake. The Y2K meltdown. The "Code Red" worm. Telecommunications. What do these things have in common? All failed to live up to their immense hype.

In the case of all-optical networking, the hype fueling much of the industry's plans to infiltrate service provider networks has long since faded, and providers are now looking for ways to meet their needs on a case-by-case basis, not on a technology basis.

Hype is dead, realism is in Lighting the way-- steps to an all-optical network

All-optical networks may have once been portrayed as the only way to build networks of the future, but the reality is far different. The more realistic approach entails finding ways for the all-optical and electrical domains to harmonize. Yet just as building a true all-optical network is years away, the incremental steps necessary to eliminate electrical elements in service provider networks will be time-consuming as well.

Although the benefits of an all-optical network are greater with fewer electrical elements, the transformation must proceed piece-by-piece as technology allows. Ironically, while so many equipment vendors initially pushed the all-optical mantra, service providers don't appear very willing to part with their faithful electronics just yet. Only as various all-optical technologies prove their viability will they earn the right to displace electrically based equipment.

Now service providers face the arduous task of determining which electrical elements to oust in favor of all-optical technologies and which must stay to perform unmatched functions. By using a phased approach, their task will be a little easier and most likely cheaper in the long run.

With demand rising, finding ways to maximize infrastructure investments means looking at and deploying new technologies. But the generally conservative mindset of most service providers makes them somewhat hesitant to adapt. With all-optical networking, clearly the best place to start is in the heart of service provider networks: the core. There, service providers are only concerned with the number of bits getting through and the rate at which they are doing so, says Ralph Bennett, president and chief operating officer of All-Optical Networks.

Working on that premise, Broadwing Communications' optical network now sports an all-optical core based on Corvis equipment. The company wanted to get rid of passive regeneration and take unit costs down another layer: “We wanted to make it easier to manage,” says Mike Jones, Broadwing's senior vice president of engineering.

So far, Broadwing's move to an all-optically switched core is paying off. The provider recently won a $180 million wholesale deal with Teleglobe and has reportedly reeled in other unnamed providers for similar contracts. In 45 days, Broadwing had installed 31 OC-192 segments for Teleglobe, and the provider also recently revealed that it delivered 38 OC-192 circuits across the U.S. in less than 60 days.

“Most of that was on the Corvis network, and the provisioning for that really took 30 days,” says Chris Nicoll, vice president at Current Analysis. “What took longer was having to go through the Nortel re-gen network.” He adds, “Broadwing has to get competitive advantage by offering bandwidth in a third of the time [it takes for other providers].”

And while time-to-market is key for turning up new services, so is reduction of provisioning time: The shorter the provisioning cycle, the greater the sales potential for the provider.

“Enterprises find it hard to order an OC-192 a year in advance,” Nicoll says. “Most aren't sure what the market will be like in a year.”

That in turn will slow customer demand unless those provisioning times can be lowered, explains Nicoll. “They can see two months out, but it's a lot harder to see 12 months out,” he says.

Others agree. “Broadwing's success in turning up wavelengths verifies Corvis' vision that the amount of [optical-electrical-optical] can be reduced,” says Ron Kline, senior analyst at RHK.

But although the Corvis equipment is working well for Broadwing, there is a lack of maturity of photonic systems products from vendors such as Lucent Technologies, Nortel Networks and Sycamore still haven't arrived yet.

In the meantime, Qwest Communications has also signed a deal with Corvis to equip its “all-optical express network.” Qwest plans to use Corvis' all-optical switching and transmission products and its dense wave division multiplexing system.

While Broadwing probably turned heads when it moved forward with its all-optically switched core, others are not idle. But just as all service providers' networks are different, their migration plans must vary just as widely. As a result, a cookie-cutter migration strategy won't work across service providers or network segments.

Other considerations: How much of a service provider's network is based on legacy switches? How much on gigabit Ethernet? And how much of the network is static as opposed to dynamic?

Essentially, service providers could implement products point-to-point and then migrate piecemeal to all-optical rings or mesh networks, maintaining their optical-electrical-optical (OEO) switches for use as grooming switches for on-ramps to the all-optical portion of their networks. The idea behind building the all-optical “express” networks is to tie together cities that generate the highest percentage of traffic in a service provider network. With traffic fed by OEO collector networks, the amount of offloading grows.

OEO is the only way to get intelligence. OO gives you no visibility in the network. The only thing you know is whether or not [traffic] gets there. --Harry Carr, Tellium

It's the fastest way to move from current OEO [networks] to satisfy growth, and we're not talking about ripping stuff out,” says Bob Wohlford, Corvis' senior vice president of marketing.

Some service providers, particularly those with legacy networks, are building completely separate overlay networks that can connect at gateway points. “This is where the characteristics of optical networks benefit service providers,” Nicoll says. “If you are facilities-based, you can easily pull up another fiber pair with completely new equipment running parallel. You can send wavelengths between systems relatively easily.”

The implementation of equipment with all-optical and OEO switch fabrics is gaining traction. “At Supercomm, we demonstrated our CoreDirector OEO switch, which is a hybrid of OO and OEO,” says John McIlwane, senior product manager for Ciena. “Express traffic goes through as all-optical or granular traffic goes through the electrical fabric. It uses the advantages of OEO and OO all in one system.

“Most service providers are excited about the all-optical architecture because they have the flexibility to use a mix of the two,” McIlwane adds. “It is critical to offer both you can't just deliver all-optical.”

The merits of an optically-switched core are clear. But looking past the core, the future of all-optical networking gets cloudy. While many of the problems are due to immature technology, for the most part the functionality provided by electrical interfaces can't currently be substituted.

“Service providers need to move to all-optical in the core immediately,” says Wohlford. “But moving to the edge is a whole different story.”

While using electronics in the core can send costs up, using them on the edge can bring them down. In addition, it can be less costly to use existing components: OC-48 accounts for 14% of metro access usage, and OC-12 and OC-3 account for more than 70% of that market, according to Bennett. “Electronics handle that nicely,” he says. “Until there is a need for it, why do it?”

Adds Nicoll, “I don't see a photon network at the edge. There is too much channel multiplexing and aggregation, and electronics are the most efficient way to do that.” He says although the edge and metro access seem to be far off for all-optical networking, the metro core may be a possibility.

OEO is the only way to get intelligence,” says Harry Carr, chairman and CEO of Tellium. “OO gives you no visibility in the network. The only thing you know is whether or not [traffic] got there.”

RHK's Kline agrees, and points out that the need to do wavelength interchange is a rough spot. It's tough to do in the optical domain, he says.

Yet the problems don't stop there: “OO needs a bunch of development in components,” Kline says. “They need to get past the re-gen problems.”

Issues such as wavelength management and the lack of standardization of GMPLS are also problematic. Reliability is a problem as well because it is typically addressed with a digital wrapper but needs electronics to be successful.

Despite the hurdles and initial hype, all-optical networking has built a solid foundation in key areas of service provider networks. That foundation is only likely to grow alongside demand for more efficiency and more capacity.

return to top

---

Hype is dead, realism is in

by Liane LaBarba

Service providers may have realized the benefits of all-optical technology in some portions of their networks, but not in all. Concurrently, the promotion of “end-to-end all-optical networks” has ended as abruptly as the slowdown in service provider spending started.

“We are passing from the age of ‘technology is king’ to focusing on what makes [the network] better,” says Mike Jones, senior vice president of engineering for Broadwing. “The next cool thing won't be rewarded over the next few years. There is a lot less excitement about throwing capital at things without returns attached.”

Daniel Flohr, CEO of FiberCity Networks, agrees: “It doesn't make sense to plug technology for technology's sake.” FiberCity is using gigabit Ethernet to deliver high-bandwidth services to multi-tenant unit buildings the economics of using electronics simply are too persuasive.

Beyond the technology issues, most agree the all-optical debate has shifted to more of an economics issue. The simplistic view of “photon good, electron bad” is gone, according to Carl Russo, Cisco Systems' group vice president of optical networking. “Service providers only care about the demonstration of significant [operational] efficiencies and the economics,” Russo says, adding that as the economics improve more all-optical technologies will make their way into networks.

“Gig-E still involves OEO conversions,” Flohr says. “It remains a hybrid of optical and electrical.”

And rather than the initial hope for the network totally freed from electronics, that hybridized approach for networks appears to be the long-term solution providing the best each has to offer.

return to top

---

Lighting the way--
steps to an all-optical network

1. Increase the maturity of optical switches
   

   The key to photonic networks. Issues such as component supplier OEM ceasing development of 3D MEMS switching fabric are problematic because that fabric is crucial to scalability

2. Transition networks to support photonic switching
   

   Networks must be able to support 2.5 Gb/s, 10 Gb/s, 40 Gb/s

3. Improve component development to usher in optical management
   

   The interchange of wavelengths is critical, yet currently possible in the electrical domain only. Components like gain compensation modules may be helpful

4. Define standards
   

   GMPLS necessitates the control of packet routing and the assignment of those packets to wavelengths. Some support an overlay model with IP routing and signaling protocols independent of protocols to control the optical layer

5. Continue innovation
   

   Metro and access portions of network still require new sources of technology and more research and development

6. Advance economics beyond the core
   

   As the variety and maturity of products increase, so will carrier demand, which will drive costs down

return to top