Adoption of transparent network architectures requires much more than just a set of good optical switch technologies. Component and system designers will have to devise practical ways of dealing with the dynamic optical paths of a transparent network.
Peter J. Farmer
Optical switching has received much attention as an enabling technology for next-generation optical networks. Reducing the number of optical-electrical-optical (O-E-O) interfaces in a network, it is said, will lead to significant cost reductions and continued progress in bringing ultrahigh-bandwidth services to corporations and individuals. Early-stage companies developing optical switching systems have received extremely high valuations from acquiring companies and the financial markets. Hence, Xros, Inc., with only 90 employees and a prototype optical cross-connect switch, was bought for $3.5 billion by Nortel Networks (Brampton, Ontario, Canada) in March 2000. In that same month, the stock of Agilent Technologies (Palo Alto, CA) shot up 43% on the day of its announcement of a "bubble jet" optical switch technology. Corvis, Inc. (Columbia, MD), a 2-year-old company promising all-optical networks, was valued at $40 billion-more than the market capitalization of General Motors-shortly after its initial public offering in early August.
Clearly, optical switching is a promising technology for provisioning, protection, and restoration of high-bandwidth services. At Strategies Unlimited, we have forecasted a multi-billion dollar global market for optical switching systems within two to three years. The adoption of optical switching technology and transparent network architectures will not be automatic, however. The reduction of O-E-O interfaces presents a special set of challenges to network designers-and a new set of opportunities for component and systems manufacturers. Only as these challenges are met will optical switching be adopted widely in long-haul and metropolitan networks.
Adaptive response required
One of the most basic issues associated with a transparent network is that the characteristics of its links are not static. A conventional O-E-O network consists of a fixed and finite set of point-to-point links. A network in the USA might consist of many individual links every 200 to 500 miles (see Fig. 1). Any one of the links between two city-pairs can be described in terms of its transmitter power level, data rate, total distance, DWDM deployment, amplifier spacing, and fiber quality, including attenuation and dispersion characteristics. Each individual link will have been engineered to allow for reliable transport of information from one point to the next.
For each of these links, network engineers will have determined what transmitters, amplifiers, and dispersion management components and systems should be used. Thus, in our example, network engineers would need to decide what type of erbium-doped fiber amplifier to place at Ft. Scott, KS, for the northward-bound link between Oklahoma City and Kansas City.
In an O-E-O network, the engineer can ignore the "true origin" of the traffic to be carried on that link and the path by which the traffic has been carried. Amplification requirements at Ft. Scott are unaffected by whether the traffic originated in Los Angeles, Dallas, or Miami, since all traffic will have been regenerated electronically at Oklahoma City. A backhoe might cut a fiber bundle near Albuquerque, NM, and divert Los Angeles-to-Kansas City traffic to another circuit (see Fig. 2). Such an event, however, has little relevance to how the amplifier at Ft. Scott is engineered.
The situation is much more complex in a transparent network. When O-E-Os are eliminated, one can no longer be sure of the quality and dispersion characteristics of a signal arriving at Ft. Scott. The path followed in Figure 2, for example, is more than 400 miles longer than that of Figure 1. Quite likely, the noise and dispersion characteristics of a signal transmitted from Los Angeles will differ over the two paths. It is likely that amplifier power or dispersion compensation levels at Ft. Scott will need to be adjusted in real time, via servo control, as path lengths and connectivity patterns change through the operation of transparent optical switches.
New devices necessary
The implication is that optical switching requires a new class of subsystems and components capable of adaptive response to changing network conditions. We anticipate that a number of new devices will be necessary for optical switching to achieve its promise.
Among these devices will be variable-output amplifiers that monitor signal strength levels to determine appropriate levels of gain, variable optical attenuators to further assist in the management of power levels, dynamic gain equalization systems that adapt as changes in optical path length alter the gain response across channels, and dynamic dispersion management systems. Development of these systems poses a great challenge, since the systems would have to determine and implement-in real time-an appropriate level of dispersion compensation as the topology of the network changes.
Extensive deployment of transparent optical switching also awaits a practical method of wavelength conversion. In an O-E-O network, wavelength conversion is an automatically available feature. Light entering an O-E-O on l1 can be retransmitted on l2 or any other wavelength. In a transparent network, by contrast, operators face the risk of "wavelength blocking." Referring back to Figure 1, for example, it could be that l1 is available between Los Angeles and Oklahoma City, but that l1 is already in use between Oklahoma City and Omaha, NE, on a route that transits through Kansas City. Consequently, l1 is unavailable for the link between Los Angeles and Kansas City.
Impact on industry
Overall, then, adoption of transparent network architectures requires much more than a set of good optical switch technologies. To achieve the promise of a reduction in O-E-O conversions, component and system designers will have to devise practical ways of dealing with the dynamic optical paths of a transparent network. Those focused on optical switching devices may find this outlook discouraging. The "flip side," however, is that there is significant opportunity to those companies that can integrate monitoring, processing, and optical component technologies and develop a new class of systems for management of signal power and dispersion levels. This need may have a profound impact on how the optical components industry continues to evolve.
Of the major players in the industry, the optical components arms of Lucent Technologies (Murray Hill, NJ) and Nortel both have extensive capabilities in electronics as well as optics. While both JDS Uniphase (San Jose, CA) and Corning Incorporated (Corning, NY) are industry powerhouses, their abilities with high-speed electronics are somewhat limited. In the coming year, it will be interesting to see what alliances and business combinations emerge as electronics and optics begin to merge and how such activity might accelerate the adoption of optical switching technology.
PETER J. FARMER is director of optical networking at Strategies Unlimited, Mountain View, CA.; e-mail: email@example.com.
FIGURE 1. Hypothetical US optical-electrical-optical network might consist of individual links every 200 to 500 miles. In this case the circuit from Los Angeles to Kansas City consists of five individual optical links, ranging in length from 250 to 475 miles.
FIGURE 2. Following a fiber cut, traffic from Los Angeles to Kansas City would have to be rerouted.