Startups tackle optical network evolution from divergent angles

Startups tackle optical network evolution from divergent angles


Most pundits agree that carriers will evolve their fiber-optic architectures away from Synchronous Optical Network/Synchronous Digital Hierarchy (sonet/sdh) equipment and toward more data-friendly systems, particularly at the long-haul network core. However, debate continues over how this evolution will take place and what role sonet/sdh will play in these future infrastructures. A pair of startups have announced their visions of the optical future--and while aspects of their philosophies overlap, they also illustrate how much disagreement still exists over the shape of tomorrow`s networks.

"Carrier customers say that the optical core could be dumb, point-to-point bandwidth, or it could be intelligent, dynamic bandwidth that could be managed better," says Charles Chi, vice president of marketing at Lightera Networks (Cupertino, CA). The optical internetworking vision promoted by large switch makers--in which gigabit and terabit switches and routers connect to the optical core via dense wavelength-division multiplexing equipment (DWDM)--is an example of the former philosophy, he believes. However, Chi concludes, "What carriers tell us is that they`re looking to build a much more manageable, dynamic, intelligent core."

Chi says that carriers` four priorities for optical network cores are scalability, flexibility, survivability, and lowest cost-of-ownership. The company has targeted these priorities through its LightWorks product line, which includes a core switching element called the Core Director, an operating system, and a network management platform called the LightWorks OpCenter. The Core Director switches would sit within the network core in place of sonet/sdh add/drop multiplexers, digital crossconnects, and the optical crossconnects now in the labs of several vendors. The new switch would accept transmission from a wide range of equipment--including sonet/sdh gear, Asynchronous Transfer Mode (ATM) and Frame Relay switches, and the new generation of gigabit and terabit Internet protocol (IP) data switches and routers--and provide a high degree of granularity, from OC-1 to OC-768.

"The bulk of the revenue, interestingly enough, is all in Layer 1 private line, particularly for the new carriers," says Chi. "The transport core needs to support a range of capacities that the services require."

In addition to scaling capacity within the switches--and a single equipment bay will accommodate 640 Gbits/sec of capacity--Lightera has designed its equipment to provide flexible routing and a suite of network protection schemes. For example, carriers will be able to choose where they want traffic to originate, where they want it to end, and the network will determine the best way to route the signals from origin to destination. Meanwhile, the network will also provide a combination of sonet/sdh point-to-point and ring protection, as well as mesh protection.

Adding new equipment such as the LightWorks Core Director to legacy networks presents a challenge for incumbent carriers. Lightera has discussed this issue with several potential customers, says Chi. Two possible schemes appear to be emerging. One would have the Core Directors comprise a super aggregation layer on top of existing sonet/sdh and DWDM equipment. Alternatively, carriers might use the equipment as the basis of separate IP-based networks for the delivery of high-speed data-based services.

The logistics of moving from current topologies to the optical networks of tomorrow were very much on the mind of Sycamore Networks (Chelmsford, MA) as it developed its new product line. Therefore, the company has focused on letting carriers decide when and where to upgrade to optical networking.

According to marketing director Jeff Kiel, Sycamore sees this evolution as occurring in stages. The first is "adaptation," which entails the determination of where a carrier needs optical networking and then introducing it. Next is "transport and switching," where carriers scale networks via wavelengths and introduce wavelength-based service provision. As the number of wavelengths through the network grows, carriers will move to the final stage, "advanced switching," where carriers will enhance the scaling and overall performance of their networks. Naturally, such a network evolution requires robust network management.

To speed this transition, Sycamore plans to take advantage of existing sonet/ sdh equipment in the core. For example, the company`s new SN 6000 Intelligent Optical Transport Node is designed to enable OC-48 service provision using OC-192 SONET infrastructure; the system plugs into the OC-192 backbones via a passive mux/demux unit within the OC-192 sonet/sdh unit.

Outside of the core, Sycamore has announced the SN 8000 and SN 8400, which provide intelligent add/drop capabilities for access, interoffice, and regional optical network applications. Again, the systems are designed to integrate sonet/sdh protection features to enable the selection of restoration options such as dedicated, shared, and priority bumping on a per-wavelength basis. As in the network core, the emphasis will be on high-speed applications.

"Right now we`re really just focused on the -12s, the -3s, and the -48s," says Kiel. "We don`t see ourselves going below that because good equipment is out there that does that well." To allow legacy sonet/sdh networks to handle the low-speed applications for which Kiel says they are well suited, Sycamore will offer the SN 860 sonet/sdh Optical Networking Adapter, which acts as a gateway between the existing sonet/sdh infrastructure and the higher-speed optical network. The adapter enables the sonet/sdh traffic to be transported at its native 1300-nm wavelength alongside the optical networking signals that will be running in the 1500-nm window.

Kiel adds that one of Sycamore`s best value propositions lies in the ease with which its equipment can be installed and configured. "If you look at network builds today, it`s not uncommon to go six months--particularly with OC-192 systems," he says. "So what we`re looking to do is bring that six-month service deployment window down to six days--and then in the future we`ll take that down to six minutes, and then eventually moving to almost instantaneous."

Kiel bases such aggressive claims on the software that controls the various boxes. When a system is hooked into the network, its software communicates with the other systems on the network. The system then measures the physical loss between the nodes to determine the plant characteristics. The system then performs a self-inventory of the cards in the network. Finally, the system software performs power balancing within network. Any time there are any adds or changes to the network--which Kiel says requires only the change of a line card--the software automatically balances the power levels once again.

Sycamore has one advantage that most of its fellow startups do not--a paying customer. Williams Communications began taking deliveries of the SN 6000 in March and expects to receive the SN 8000 beginning this month. According to Wayne Price, director of network architecture at Williams, the carrier has deployed about 67 Ascend GX 550 switching systems nationwide that come equipped with OC-48 trunks. Rather than connecting them directly to its Nortel Networks OC-192 muxes, Williams will use transponders to link the Ascend switches into Sycamore equipment as a way to turn up OC-48 services.

Williams believes that an intelligent gateway between the large routers and the optical layer will be a necessity--even as router manufacturers attempt to integrate DWDM functions as well as sonet/sdh interfaces into their equipment. "One of the things that you`ll hear from a lot of router manufacturers is that they want to start pushing DWDM components into the routers. We don`t believe that that actually scales very well," Price says.

"You can do that maybe internally in a network, if like the carrier can control the router and the DWDM infrastructure, then it doesn`t seem to be a problem. But to take, say, a wavelength-specific interface from a router and stick it in a WDM mux/demux would be very dangerous because if the customer decided to hit you with hot power or the wrong wave, he could actually knock down traffic on adjacent waves."

"I think you`re going to start seeing carriers beginning to implement optical networking this year," Kiel predicts. "Typically it`s going to be the OC-48 speeds and OC-48 services, really focused primarily on serving the needs of ISPs that need to interconnect their routers and interconnect their data networks. And then over time you`ll start seeing that then move down in speed to the -12s and to the -3s." q

More in Network Design