The announcement by Lucent Technologies Inc., Murray Hill, N.J., in November that it had developed a true optical cross-connect using micro-mirror technology (see news) was like a bombshell going off in the optical networking world, shaking competitors and network operators alike.
While the Lucent development is a significant advance, the company has not magically metamorphosed its products to the seamless, smart all-optical network of the future. Much remains to be done technologically to complete the transformation to an all-optical world, and there are still many opportunities for the company's competitors to provide intelligent optical solutions.
Still, the Lucent announcement could be the spark that ignites widespread adoption of intelligent optical network systems and topologies, even if some of the pieces are missing. Intelligent optical systems will still open the door to making these systems more flexible, more capable and, above all, much less costly.
Intelligent optical networks are core optical networks using dense wavelength-division multiplexing (DWDM) technology, in which whole wavelengths, or lambdas, are switched, rather than breaking them into their component data flows. The waves carry high-speed data, typically at OC-12, OC-48 or OC-192 levels. That switching is accomplished by converting lambdas to electrical signals and switching them in a huge digital cross-connect. The hope has been that a true optical switch could be employed in the future to switch waves without an electrical conversion.
Other salient features of intelligent optical networks are that the hubs where the digital cross-connects are located are intelligent and are in communication with other hubs in the network so they can find optimum routes for traffic and instantly reroute traffic where there are outages. Network topologies can be ring or mesh. It is possible to use a mesh topology and still have the protection of rings because the intelligent hubs provide instant rerouting.
Mesh topologies are desirable for several reasons. They do not require half the bandwidth be reserved for restoration, as synchronous optical network (SONET) rings do. Large cross-country networks with multiple connected rings sacrifice large amounts of bandwidth for protection. Without the ring architecture, it can also be possible to leave off SONET equipment altogether and modulate services, such as Internet protocol (IP) and asynchronous transfer mode (ATM), directly onto fiber. The networks rely on the intelligent mesh to protect service. This can further improve bandwidth utilization and save the costs of SONET.
The magic that service providers are hoping for from intelligent networks is improving cost and time to market. Deployment of a new services and ramping up capacity in a traditional core network requires truck rolls to install expensive equipment, including pricey regenerators. This costly process can take months. Even without optical switching, intelligent networks greatly speed the process, requiring only that new network cards be added at each end of the service.
The other magic is the ability to scale networks to heretofore unimaginable levels. Without optical switching, a service provider would have to buy very expensive digital cross-connects or set up a web of optical fibers that are patched manually, both of which are unworkable. "We are looking ahead to Internet scale, very rapid doubling and exponential growth curves," says Mike Guess, vice president of engineering, IXC Communications Inc., Austin, Texas. "And we don't have to look very far ahead to see thousands of OC-48s, so we have to make the technical decisions today to get there."
Intelligent Vendors
Several large vendors and startups have announced intelligent optical technology, though very little is currently deployed by service providers. An exception is Williams Communications Inc., Tulsa, Okla., which has announced it is testing and plans to deploy Linthicum, Md.-based Ciena Corp.'s MultiWave CoreDirector product. Williams will use the product to provide an ATM service that will carry voice, data and video in the same network.
Williams chose to use an intelligent optical approach because it is able to scale to terabit levels and provides rapid provisioning of services, often without truck rolls. MultiWave Core is the first product to grow from Ciena's purchase of Lightera Networks Inc. in early March 1999.
Monterey Networks Inc., Richardson, Texas, which was acquired by Cisco Systems Inc., San Jose, Calif., on Sept. 29, coined the term "wavelength routing" and was one of the first companies to throw its hat into the intelligent optical ring. The early iterations of its technology do not use optical switching, but rather employ two 256-by-256 digital cross-connects (at the OC-48 level) to route waves. The company has not yet announced any large purchases of its equipment.
Along with Monterey, Sycamore Networks Inc., Chelmsford, Mass., was in the first wave of companies to announce wavelength router-type products, and like Monterey, its technology was designed to be all-optical-ready--it used electronic cross-connect technology but could accommodate optical when available. The company made a sale of its wavelength-provisioning product, the SN 6000, to Williams Communications in March 1999, when the product was launched. Nov. 9, the company announced the rest of its intelligent optical system, the SN 8000, which includes modules of OC-3and OC-12 waves. This makes intelli-gent optical switching available for lower-scale applications.
The company also allows mixing of services on a wavelength, so a carrier does not have to devote an entire wave to just one service, such as ATM, if that service is only lightly subscribed.
Nortel Networks Inc., Richardson, Texas, has the Nortel Optera Connect product, its first pass at intelligent optical networking, and it is geared to support Nortel's Optical Internet initiative. It includes terabit routers that do native IP switching and/or more demanding multiprotocol label-switched switching. The Nortel product is distinctive in that it can switch whole waves at their full capacities of 40 gigabits per second (gbps), 10gbps or 2.5gbps, or can switch fractions of those waves. Louis Paré, director of high-capacity networks, optical networking, says, "the advantage of being electric over photonic switching is that we can switch fractions of large streams."
Corvis Corp., Columbia, Md., which bills itself as "the world's first all-optical communications network," has yet to reveal the technology behind its meshed intelligent network system. The company had a splashy debut at the Supercomm '99 convention in Atlanta, but has shown little of its technology since then. Corvis announced in the fall it is increasing the capacity of its optical networking products to support multiterabit IP networks, and adding interfaces for OC-192 and OC-192 concatenated systems.
Other players waiting in the wings include Corvia Networks, Sunnyvale, Calif., a provider of multiple-terabit-per-second switching systems, which has not yet revealed its technology. Qtera Corp., Boca Raton, Fla., is a startup that will offer a network system with photonic switching. The company, which has not announced its technology, will debut its products in March 2000.
New Technologies: Smarter, Faster, Farther
The Lucent optical cross-connect is a critical product that will make large wavelength-based optical networks more feasible and capable, and probably more affordable, than ever before. The product is a matrix of 256-by-256 waves, which are switched by microscopic mirrors that bend with changes in heat. In addition to functioning as a cross-connect in an optical hub, the product, which is a one-inch chip, could be used as an optical add/drop multiplexer (OADM).
However, other technologies will be needed if the product is to be called an optical router, says Chris Nicoll, carrier infrastructure analyst with research firm Current Analysis, Sterling, Va. "Lucent needs to add the ability to recognize the WaveWrapper technology it debuted earlier this year, which provides optical routing and addressing information that it can act on. Then it becomes a router." WaveWrapper adds routing and prioritization information to a wave.
The second technology needed for all-optical networking is wavelength conversion, Nicoll says. Unless a switch is able to change the color of a wave, it will be constrained in the amount of input and output it can handle. If a switch wishes to send a wave to Houston and that color is already occupied by traffic of the same color coming from another fiber, manual provisioning may be required or there could be traffic interference. A short-term solution could be to use an electrical cross-connect with the ability to convert wavelengths, Nicoll says.
Lucent has said the cross-connect will provide cost reductions of 50 percent to 75 percent in DWDM optical hubs. Guess says true optical switching will be valuable only if it can hit those price points. "Without true optical, I would rather have electrical at the central stage. With electrical you can regenerate the signal, and it could be more cost effective than running it through an optical switch and doing regeneration as a value-add. Plus, with electrical I can change waves." Also, if the spans that can be put between regenerators become longer, it becomes increasingly cost-effective to do optical switching.
Indeed, optical networks are due for large increases in the distances between regenerators, a key factor in reducing costs. New laser technology improvements, combined with forward error correction (FEC), are increasing the power of transmitters. DWDM systems typically can run up to 50 miles on their basic power budget. In 2000, that is expected to increase to 100 miles and to as much as 400 miles. The effect would be to make an area of several large cities, such as on the East Coast of the United States, potentially a metro system.
Sycamore has already achieved spans of 1,600 miles, and Corvis at least 1,000. Sycamore, Qtera and other unannounced startups in the intelligent networking space are said to be working toward 7,000 miles without regeneration, which would enable coast-to-coast networks in North America without regenerators. Besides better lasers and forward error correction, the ability to stay all-optical will also be a factor in eliminating regeneration.
If waves can be added or dropped optically, rather than having to take the signal to electricity as is done today, then there would be less power loss in the network. The Lucent optical cross-connect technology could be used as an OADM.
Also improving are technologies to separate the different colors, or wavelengths, of DWDM. Improvements in grating technology--which works like a prism--have helped increase the number of waves from just 16 to as many as 40. Simultaneously, filtering on waves will enable more accurate access to waves at OADMs.
