In the swift-moving world of optical networking, long-haul operators are finding a new generation of technologies to facilitate the rocketing demand for higher and higher levels of bandwidth.

Among them is micro-electro-mechanical systems (MEMS), an optical switch technology that employs microscopic mirrors to route signals from fiber to fiber in a network cross-connect or node. Theoretically, the essentially all-optical switch bypasses the need to convert optical signals to electric signals, as is done with traditional switches.

One of an array of 256 microscopic mirrors, each the size of the head of a pin, tilts to steer lightwave signals from one optical fiber to another in Lucent Technologies' WaveStar LambdaRouter.

Nortel Networks' Xros division and Lucent Technologies have both introduced MEMS-based switches. With long-haul providers, including those with Internet backbones such as WorldCom, faced with dramatic increases in bandwidth demand, the horse race to develop and bring to market optical switching technologies is on.

In addition to MEMS, other advanced optical switching technologies exist, including an intriguing technique developed by Agilent Technologies Inc. that uses a combination of inkjet and planar lightwave circuit (PLC) technologies that re-direct light without the help of mirrors. Agilent's switch is composed of a vertical and horizontal array of fixed aligned waveguides. Light is transmitted across a horizontal path from the input to output port until a switch command is issued. When commanded, a bubble is created at the intersection of the appropriate waveguides and the light is reflected down a vertical path to the switched port. This bubble is formed using the same technology used in inkjet printers.

Agilent's technology has been adopted by Alcatel for its recently-announced CrossLight photonic cross-connect.

But in the eyes of one network provider, WorldCom, MEMS is showing the most promise in the near term. "MEMS is the first one that looks like it meets the basic requirements," says Jack Wimmer, vice president of network technology and planning for WorldCom Inc. Those requirements include scaling to large numbers of ports, relatively low loss and attenuation throughout the system, and support for a wide spectrum of bandwidth.

Wimmer, who stresses that WorldCom Inc. prefers to not specify core technology to its suppliers—instead focusing on system level technology, says, "Our view is that MEMS appears to have great promise for large, scalable, truly photonic cross-connect systems. It looks like it will win the initial horse race compared to other technologies that we're looking at."

WorldCom, according to Wimmer, has tested MEMS as a switching technology on prototype systems. "We have been looking for technology that will provide us a very scalable photonic cross-connect system for a number of years now, and we are fairly optimistic that MEMS is going to provide at least some of the initial solutions in a relatively short time frame," says Wimmer.

The Williams Communications' Multi-Service Broadband Network.

The Xros system, according to the company, is capable of connecting, in milliseconds, up to 1,152 pairs of inputs and outputs with a transparent all-optical pathway. The Xros cross-connect consists of two facing six-inch-by-six-inch arrays of 1,152 mirrors each. Incoming beams of light are directed to their destinations by these movable mirrors. Thus, an entire incoming fiber optic channel can be switched as a whole to any output port.

The Xros scheme also facilitates redundancy by allowing network planners to create an architecture where multiple alternate paths can be made available to protect each link.

Xros' optical switch using MEMS.

The ability to switch traffic photonically frees a long-haul carrier from worrying about the protocol and bitrate of network traffic, says Greg Reznick, president of Nortel's Xros division. Reznick says that with major backbone traffic becoming more datacentric, the parameters for high-quality delivery have changed. While entire networks used to be optimized for voice, with Sonet often the preferred means of transport, the emergence of data (including packetized voice), means "there's got to be ways to improve efficiency and density of networks."

With networks becoming increasingly heterogeneous, protocol independence, or transparency, becomes more important. For example, instead of mapping Ethernet into a Sonet transmission, Reznick says it's important to simply send Ethernet from one place to its destination. By relying on an all-optical pathway at the cross-connects, network operators have more flexibility in handling different types of traffic and provisioning new network clients faster, Reznick says.

"The technology is changing so fast in the optical space," Wimmer points out, "that if we can truly get to that transparent model in a way that works, we think it will have a huge benefit in a relatively short period of time."

WorldCom today is "deploying almost all of our capacity at 10 Gigabits per second (Gbps)," says Wimmer, who adds, "but we are going from tens of wavelengths to literally hundreds of wavelengths over the next six to eight months." Wasting no time in moving beyond that threshold, the company is testing 40 Gbps per wavelength technology in the lab, and Wimmer expects it to become deployable with the next year or so.

And with the Institute of Electrical and Electronics Engineers (IEEE) finalizing standards such as 10 Gbps Ethernet, which is slightly different than the bitrate and format of OC-192 Sonet, "we can very easily see a time not too far down the road where the wavelengths in the network carry various combinations of these signals," says Wimmer.

Port provisioning ease further enhances MEMS switches because of the inherent transparency of the system. Today, "it takes months to get something provisioned," says Reznick. While looking for a machine that will perform provisioning and restoration of the network, Wimmer adds that it's important that the switch "be one where we don't have to be explicit about what ports we configure" or know in advance what wavelength is going to carry which signal.

Wimmer says that as WorldCom evolves its network from trunking at 10 Gbps to trunking at 40 Gbps, it's important that the company doesn't have to replace its existing infrastructure as part of that conversion.

As an example of how quickly WorldCom is expanding its network, which includes the UUNet Internet backbone, Wimmer says that last year, the company installed as much capacity in the network as existed at the time of the MCI/WorldCom merger—which took those two companies 18 to 20 years to build. This year, WorldCom will install that much bandwidth in nine months. Next year, it will install that much bandwidth every month. That bandwidth will be managed in large aggregate pipes at the wavelength level, rather than at traditional DS-1 and DS-3 levels. "That's why the (optical switch) machines are such an important initiative for us," says Wimmer.

Lucent recently joined Nortel in the MEMS switch space with its WaveStar LambdaRouter, an all-optical switch that Lucent says is capable of routing 10 times more information than is traveling today across the worldwide Internet. It uses 256 micro mirrors which reside on less than a square inch of silicon. In April, Lucent announced that Global Crossing Ltd. will be the first carrier to deploy the Lambda-Router, together with its 320-channel Dense Wave Division Multiplexing system.

The other side of the coin

Despite the momentum built by MEMS technology by its adoption by Nortel and Lucent, MEMS hasn't won over everyone. Andy Wright, chief technologist for optical networks for the Williams Communications Network, says that MEMS "is a viable technology and a good way to do a scalable switch solution." But he adds that early implementations of MEMS-based switches have "fairly high" loss associated with passing a signal through a DWDM demultiplex system, through an optical switch, and through a DWDM multiplex system.

While he wouldn't disclose how much loss Williams encountered in its evaluations of MEMS technology, he says the loss level "is going to force us to stay with an O-E-O (optical-to-electrical-to-optical switch) architecture in the near term."

Wright also expresses concern about relying on a single vendor for MEMS-based optical switches, and cites a lack of technological compatibility in the MEMS optical switch market. Additionally, he says the reliability of MEMS mirrors to switch accurately and rapidly, perhaps many thousands of times per day, is an issue. The failure of a single mirror could mean the replacement of a large portion of the switch fabric, or the entire switch fabric, Wright says.

Alcatel's Mike Jamgochian, senior product line manager for the company's CrossLight photonic cross-connect, says that the company chose Agilent's bubble technology over MEMS because of time-to-market issues. Agilent's technology was available more quickly and is scalable to larger numbers of ports. Alcatel hopes to announce beta tests of its CrossLight product next March with several carriers. (The Xros MEMS-based switch will be ready for production availability in January.)

However, Jamgochian says, "We continue to look at MEMS because of its ultimate scalability . . . We believe the maximum size over the next two- or three-year period is probably 4,000 ports." He didn't rule out the possibility of Alcatel adopting MEMS, possibly as three-dimensional-moving MEMS mirrors become available in the second half of next year.

While MEMS technology is today geared to long-haul fiber networks, as operators drive fiber deeper and as high-bandwidth applications become omnipresent, MEMS may be adapted for both access and metropolitan networks.

Reznick says that MEMS technology will become applicable to smaller access networks as it begins to offer more high-bandwidth services. As an example of how data service penetration can affect core network capacity, Reznick says a one percent increase in DSL customers translates to a 100-fold increase in traffic at the core.

Working with smaller implementations of MEMS technology is Chuan Pu, senior member of technical staff for AT&T Labs' lightwave networks department. Using MEMS-based switches in a cross-connect of 16 input and 16 output ports, Pu says the switching technology could be useful in hybrid fiber/coax networks, particularly in the fiber-based return channels. Advanced MEMS mirrors could help implement a passive model, without using active lasers, on return paths, says Pu.

"Clearly, MEMS has some interesting potential," says Paul Connolly, VP of marketing and network architecture for Scientific-Atlanta, who adds that S-A hasn't developed products based on the technology yet. S-A has invested in several optical networking companies, including Bookham Technology, an optical integrated circuit developer, and Alloptic Inc., a developer of passive optical network technologies.

The importance of new optical technologies, says Connolly, is "getting the cost out of systems." He adds that "a lot of long-haul technologies don't migrate to access network technology."

If and when cable operators turn to heavy-duty optical switching, whether to augment high-traffic in large metro systems, or to route traffic from one large cluster to another, there should be mature technologies available to meet the demand.