Bursting the Optical Switching Bubble

Long the province of academia, optical burst switching is making its commercial debut this fall as Matisse Networks begins shipping its EtherBurst optical switch. The 3-year-old company surprised some experts in the industry by emerging from stealth mode in September to announce its new product, the first using optical burst switching techniques.

Matisse says the EtherBurst optical switch is in beta trials with Lawrence Livermore National Laboratory and is expected to be in trials with a CLEC and an ISP yet this year.

Optical burst switching overcomes an increasingly common dilemma for service providers — the need to scale Ethernet beyond 10gbps without sacrificing its flexibility by running it over pre-provisioned, point-to-point optical DWDM circuits, says Mark Showalter, director of marketing for Matisse Networks.

In contrast, he says, optical burst switching combines the any-to-any flexibility of Ethernet switching with the bandwidth of DWDM. It employs optical burst transponders that switch packet traffic in bursts over different wavelengths, enabling them to be used for different colors at the same time, rather than be provisioned statically one color at a time as with DWDM.

DWDM vs. OBS In configuring a DWDM network, the number of circuit transponders is equal to the number of nodes times the number of nodes less one or N(N-1). So, a four-node network would require 4(4-1) or 12 transponders. In contrast, a four-node optical burst switching network using Matisse Networks’ EtherBurst requires only one transponder per node, or four transponders.

“With the EtherBurst optical switch, we eliminate the need for optical circuit provisioning,” says Showalter. “Essentially, you plug in our components and the optical layer automatically configures, and you don’t need to worry about what colors are where. All you are worrying about is provisioning an Ethernet switch.”

This fundamental architectural change has significant opex and capex implications for service providers, analysts say.

“It is disruptive. It is a great application, and it is the direction that carrier networks are going in,” says Michael Howard, principal analyst and co-founder for Infonetics Research, referring to Ethernet/Optical architectures deployed in support of IPTV, VoD, storage networks and more.

“The technology is extremely compelling,” says analyst Ray Mota, chief strategist for Synergy Research Group. The benefit of being able to make circuit technology dynamic and “almost work like an Ethernet switch, it’s amazing,” he says, noting “extreme” operational cost advantages. “If this works, it’s going to be great. It basically will make an optical network ... work like a plug-and-play scenario, which will reduce the total cost of ownership.”

Both analysts confirm that Matisse is the first to market with an optical burst switching product. Mota adds there are at least two other manufacturers working on systems, but he was unable to disclose their identities. Howard says that while Matisse is out of the shoot early, it is likely to be followed and could be a good acquisition target for the larger OEMs once it’s proved itself with a customer win.

Matisse Networks’ PX 1000 Photonic Node

Matisse Networks' SX 1000 Ethernet Service Node

The most recent Workshop on Optical Burst/Packet Switching — an event that’s been held since 2000 — offers some hints at other endeavors. Held in October, in San Jose, Calif., the forum, which typically features presentations from university researchers worldwide, included keynotes from Matisse and Takeo Hamada, director of IP networking research for Fujitsu Labs of America. Hamada discussed optical burst transport (OBT), a joint research project between Fujitsu Labs of America and Photonic Network Research Laboratory of Stanford University. OBT is based on work in optical burst switching and is a WDM ring-based optical data transport with electronic token-based control plane akin to FDDI.

Analyst Mota says the holdup with optical burst switching has been making an optical device work like an Ethernet switch. “The part that’s a little tricky is how to make Ethernet burst into the lights properly using a C-band wavelength,” he says. “That’s been some of the issue. [Matisse] figured that out.”

Indeed, Matisse Networks’ patented 10gbps “Tango” optical burst transponder is capable of tuning to any wavelength in the ITU C-band in nanoseconds. It also addresses some common challenges — burst scheduling, collision avoidance and QoS — with optical burst switching using its MeshWave packet processor.

“Scheduling of the optical bursts into the fiber is one of the problems the academic community has been trying to solve for some time,” says Showalter. “Our patents address our solution for bursting without collision into the optical domain.” Other proposals, he adds, assume some level of collisions will occur.

Both Tango and MeshWave are contained in an optical burst switching interface card called TAP that is included in Matisse Networks’ EtherBurst SX-1000 Ethernet Service Node.

“TAPs can communicate with all the other SXs on the network optically,” explains Showalter. “They essentially receive the packet, figure out which color is the appropriate destination color, tune the laser to that color, fire and move on to the next color.”

The SX-1000s support up to 48gigE or four 10gigE interfaces and connect to the PX-1000 Photonic Nodes, which provide the fully automated, all-optical photonic layer. Up to 32 SX-1000s — each with two 10gbps TAP interface modules — can be attached to the dual-ring photonic layer, yielding a total network capacity of 640gbps.

Significantly, while Matisse has built the optical burst transponder, it has not built the components, enabling a lower cost entry point. Prices start at $86,000 for the SX-1000 and $58,000 for the PX-1000. “What we have done is taken a standard JDSU laser off the shelf ... so we can ride the cost curves. We have taken a standard laser and put our own intellectual property and analog circuitry around it and put it into an industry standard form factor,” Showalter notes.

He adds that because an optical burst transponder can fire and receive from all locations on the ring, only one transponder is required per location, reducing the number of transponders you need to scale an optical network. “We typically need about half the number of transponders and router ports as an existing circuit DWDM network requires,” he says, noting that transponders — at $50,000 to 60,000 each — are the primary cost of building an optical network (see diagram above).