Optical switches are coming--but when?

July 1, 1998
13 min read

Optical switches are coming--but when?

Optical switching elements will provide the key to all-optical networking. But when will they appear?and when will they be deployed?

Lee Branst Caracal Communications

Switching for fiber-optic networks has become a topic of great interest and debate within the telecommunications industry. A wide variety of companies with varying pedigrees and experience in fiber optics are either introducing switching equipment specifically designed for fiber-based applications or adding optical capabilities to products originally designed for other media. This article examines what companies working in the fiber-optic switching field say are the major product, technology, and application trends for fiber-optic switching systems. As you`ll discover, there is a major emphasis among this group on the new generation of all-optical systems.

Basic issues

"There are three fundamental issues regarding the migration to all-optical switching," according to Serge Melle, director of marketing for optical networks at Nortel (Atlanta, GA). "These are: 1) What are the business drivers for optical switching? Why would a carrier want to deploy it? 2) What are the optimal architectures that will give network operators solutions to their business drivers? 3) What control and network management issues about optical switching must be resolved?"

Melle believes the answers to these questions will determine the pace and type of all-optical integration. Specifically, he thinks the introduction of hybrid-rate data services will start appearing in networks later this year. "But it will be 1999 to 2000 when the penetration rate is high enough for the move to OC-192 data services to start occurring. This will include optical-layer protection and bandwidth management," he asserts.

An optimistic outlook is provided by Pawan Jaggi, manager of the Optical Networking Group at Fujitsu Network Communications (Richardson, TX). "Extremely high-capacity, dwdm [dense wavelength-division multiplexing] systems using OC-192 [10 Gbits/sec] and OC-48 [2.5 Gbits/sec] are being deployed today," he points out. "The next obvious step in exploiting the photonic layer is the use of configurable oadms [optical add/drop multiplexers] and oxcs [optical crossconnects]." Jaggi says both these types of products are being introduced now. "The oxcs will route multiple wavelengths at OC-192 and OC-48 dynamically in the optical layer to provide huge network efficiencies. These oxcs will be deployed in mesh and linear add/drop ring networks."

As to the network timetable, Jaggi replies, "Optical switching will come in the long-haul networks initially because the use of dwdm is quite extensive in these long-haul routes already, and it makes sense to route these wavelengths dynamically in the optical domain. Once dwdm systems get deployed in the metro environment in large quantities, optical switching will find a place in these short-haul and interoffice applications." He feels that the cost of optical switching will not be justified in an enterprise or campus-to-campus network for a long time. However, he notes, "Never say never."

Robert Pullen, vice president of engineering and marketing at Tellabs (Lisle, IL) points out that today the different applications for optical networking include long-distance interoffice facilities and submarine cable networks. He feels the next market to be enabled will be metropolitan point-to-point applications between two hub offices. "Thereafter will come metro rings for connecting major hub offices together," he adds. "Then optical networking will make its way into access in a more aggressive domain."

Pullen has definite ideas on the timetable for optical switching. "By the 1999 time frame, we will have the capability to add/drop wavelengths optically on a software-configurable basis. By the year 2000, solid-state oxcs will be able to switch full fibers independently of the number of wavelengths. By 2002, we will be able to switch wavelengths on an integrated oxc basis. That will exhibit some blocking probabilities. By 2003 we will see wavelength interchange in oxcs evolve, i.e., the capability to switch wavelengths on the fly, of doing wavelength conversions and implementing maximum grooming and filling or switching capabilities" (see Fig. 1).

Slow and steady

John Fee, an optical networking veteran who is now chief technology officer at Avanex Corp. (Fremont, CA), predicts, "The migration to optical switching will come first in small switches with limited functionality. These early systems will be very coarse, limited in capability, and expensive. But as volumes go up and users start looking at different switch matrices, they eventually will become widespread."

It will take carriers with a willingness to experiment to lead this evolution. "The shift may begin in small startups with everything to gain and nothing to lose, or possibly in aggressive, high-tech carriers who want to demonstrate the technology," Fee says. Still, he notes there will be interesting developments to follow. "The questions are, Will the interexchange carriers or rbocs [regional Bell operating companies] do it first? Second, will it happen in the U.S. or overseas first?" he offers.

Seconding this gradual approach is Jeff Santos, business development manager, Optical Networks Group, Alcatel (Richardson, TX). "There will be an evolution of all-optical switching products, i.e., all-optical crossconnects," Santos says. He thinks this trend will hit the long-distance applications first, "where the granularity is the wavelength as opposed to today, where the granularity is the DS-3, DS-1, or even DS-0. I don`t see 40-plus-channel wdm applications in metro and short-haul networks in the near future. They will be in long-haul applications relatively soon. Then we will need the ability to manage those wavelengths."

Santos thinks the first application for oxcs will be to replace the fiber distribution panel "that`s doing nothing more than switching wavelengths from one input port to another output port. If it is just replacing an optical patch panel, that type of crossconnect can be somewhat lossier because there are wdm terminals, oadms, etc., around the switch. Therefore, those products are managing the optical span, not the crossconnect."

According to Santos, there will be a time frame of at least two years before there is wide-scale deployment of a switch with at least 128 ports. The next generation of oxcs will include wavelength conversion and switching--in other words, they will accept an optical signal of one frequency at one port and change the frequency of that signal on the fly before adding it to a different outgoing fiber. "This would require some kind of regeneration to maintain optical line signal standards," Santos says. "oadms are farther out on the horizon. Now we have the first generation of wdm-based products that can efficiently add/drop wavelengths. These are really terminals that also can add/drop wavelengths. However, true oadms must be software-reconfigurable like the sonet adms of today. It is probably two years out before we have wavelength converters, tunable filters, space switches, etc., that are economical, small, fast and not too lossy to have a fully reconfigurable oadm."

A word of caution is injected by Stephen Montgomery, an analyst with ElectroniCast (San Mateo, CA), a market consultancy for the communications industry. "Some observers feel that the all-optical network will never happen. There always will be some electronics involved because of the need for power." In addition, he believes that optical networks generally will not replace old equipment but will be used in new builds. "Of course a carrier will switch out a switch if a customer wants it that way for a particular business park, university, etc.," he adds. Montgomery also cautions that there will be software problems related to fiber management.

Another cautionary note is sounded by Scott Wilkinson, account marketing manager, Hitachi Telecommunications (Norcross, GA). "We won`t see long-haul all-optical networks soon because regeneration will have to be done electrically for the near future," he says. He agrees with Montgomery that old cable will remain as is because it is too expensive to replace. "But new entrants, who have doubled the amount of cable in the ground, are not wedded to the old technology," he notes.

As to the trends, Wilkinson says, "In the next five years we will see more of the same trends, like moving to higher speeds and going over wdm. We will start seeing adms and a bit of oxcs, primarily for wdm systems, because they can be protocol-independent. In the next five to ten years we will be going to all-optical systems." The reason for this time frame is that we are missing what Wilkinson calls the Holy Grail of all-optical networking, an all-optical 3R regenerator which is protocol- and wavelength-independent. "Once that appears, there will be nothing holding back optical rings. Optical rings/switches in the next two to five years will be great in certain situations, but they will not be ubiquitous," he predicts.

Along with regeneration, another necessary ingredient is all-optical wavelength translation. "Most lab models now are protocol-dependent and not wavelength-independent. When those two things come about, we will see the biggest explosion in all-optical networking. Before that we will have to be careful how it is used," he says.

Wilkinson points out that many vendors are coming out now with electrical adms that look like crossconnects and vice versa. He feels the same thing will happen in the optical area. "At some point when wavelength selection is put into an oxc, it starts to look like an oadm. Everything will be managed in one box, so we will have an oadm/oxc. Thus, you can bring any wavelength in on any fiber and take any wavelength out on any fiber. We`re working hard on this concept, but it`s five to ten years off," he says (see Fig. 2).

Mindful of predictions that are often too optimistic is Tom Canavan, director of business development, Switching Div., Public Networks Group, nec America (Herndon, VA). "On the switching side, it`s obvious to us that atm [Asynchronous Transfer Mode] is the only means able to reach the broadband multimedia goal of integrating data, video, and voice in a high-quality, public-type vehicle," he says. "When we introduced atm with a 2.4-Gbit/sec switch six years ago, the industry said, `We need voice, video-on-demand, SMDS [switched multimegabit data services],` every application you could think of. `We need OC-192s and big atm switches.` But it didn`t materialize. Instead change came slowly, because there will be no big push to convert unless there are sound economic reasons to do so."

Therefore, he believes, "Voice applications will happen in atm pretty quickly. However, data applications, such as frame relay and IP [Internet protocol], have been forcing the size and design of atm switches." But watch out for video, which Canavan calls "the sleeping giant."

Ken Garrett, director of strategic accounts, jds Fitel (Boulder, CO), believes that the initial implementation of optical switching is easiest from a disaster-recovery point of view in long-haul networks because there are recovery methods that could be integrated easily. "However, the overall market will tend toward metropolitan network applications," he predicts. He feels that the driving force in the next three to four years will be restoration. "Protection switching by itself is not enough cost-justification for a stand-alone optical-switching system."

Garrett believes that optical switching has two basic applications: physical network management and wavelength management, such as wavelength grooming. To Garrett, physical network management means disaster recovery and route diversity. The latter is the ability to load-balance traffic within a network. "Physical network management will be of primary concern for the next three to five years. It probably will be that time span before field trials start. Wavelength grooming will come in five to ten years," he says.

Paul Segre, vice president, Integrated Multi-Rate Transport Node Div., DSC Communications (Plano, TX), recently acquired by Alcatel, says, "Rates at which bandwidth is being managed are moving up, so a huge number of wideband crossconnects have been deployed. Now we are seeing broadband crossconnects beginning to take off. At the same time, wdm is entering the market at the high end." He says the early adopters for both technologies have been interexchange carriers, with rbocs somewhat slower.

However, he feels that wdm will be economical only at OC-48 rates for the foreseeable future. Thus the same will be true for oxcs. "I can imagine scenarios where optical switching would occur at the OC-3 [155-Mbit/sec] or OC-12 [622-Mbit/sec] level, but I don`t foresee that happening anytime within my planning horizon," he says. "I see broadband crossconnect DS-3 [44.736-Mbit/sec], DS-1 [1.544-Mbit/sec], OC-3, OC-12 types of rates feeding into fiber-optic terminals with wdm optical switching managing the OC-48 level as far as crossconnection. It will be an extraordinarily long period of time before the minimum level of granularity to be managed is at the OC-48 level. oxcs will manage at the wavelength level."

It`s a broadband world at present, according to Segre. "Architectures will have broadband in the middle for the next seven to ten years," he states. "These will provide various hubbing and grooming functions and be put out directly over either wdm or optical-switching networks for long-haul transport. Optical switching will be responsible primarily for wdm efficiency." He predicts that reliability services will be integrated next. "I can imagine the interface off the broadband will be unprotected, and it will have the optical network provide some 1 ¥ N protection."

Our final industry viewpoint comes from Rod Alferness, chief technical officer, optical networking group, Lucent Technologies (Warren, NJ). He keys in on a merger of crossconnect and adm functions. "There are already simple fixed adms in long-haul networks that don`t require any switching element," he says. "In metropolitan applications there will be a need for programmable or configurable adms that are smaller versions of current versions." Thus, he feels the adm will be the key to metropolitan applications. "On the other hand, wdm is finding its way into the long-distance market first. Therefore, in terms of applications of oxcs, particularly the larger ones, they very likely will be in long-distance applications first," he adds.

He opines that the value of oxcs is that they can crossconnect signals of different formats and bit rates in a single switch. "The challenge of oxcs in the long-distance market is the ability to scale them to sizes as large as 256 ¥ 256 or even as large as 1000 ¥ 1000," Alferness cautions. "I think the carriers have been convinced that they have to deploy crossconnects that work at 2.5 Gbits/sec. In contrast, the electrical crossconnects of today are 45-Mbit/sec crossconnects. If it turns out that the technology cannot be found to do that optically in a scalable way, then we will be looking for ways of doing that electrically."

oxcs are primarily useful for grooming traffic and rapid restoration, he believes. "In the metropolitan environment, the challenge is to make it cost-effective, for example, to carry a 150-Mbit/sec signal on a single wavelength and not multiplex it up to any higher rates. Crossconnection at OC-48 rates, whether purely optical or a combination of optical and electronic, will happen in the next two to three years from a commercial point of view," he states.

A look at the future

Which of the above scenarios will be correct, which will need modifications, and which will be scrapped completely? Only time will tell, of course. But the journey, as always, will be interesting, particularly as a new group of companies, those with a background in data networking, attempt to bend fiber-optic technology to their purposes. As to the optical-networking strategies announced by some members of that group, Tellabs`s Pullen says, "I believe optical networking is being hyped today more than its actual capabilities. [Some companies] would like to believe that wavelengths of light will supplant atm, sonet and sdh. I don`t think that`s true because of the ubiquity of sonet and sdh, their end-to-end network restoration and integrated network-management capabilities."

Regardless of Pullen`s feelings, those companies will have their say--and soon. u

Lee Branst is a freelance writer and founder of Caracal Communications (Redondo Beach, CA).

Sign up for Lightwave Newsletters
Get the latest news and updates.