TeraBurst's OMS A Scalable Optical Switch - OMS 2100 and OMS 2200 network switches from TeraBurst Technology - Product Information

Telecommunications, August, 2001

www.teraburst.com

One way to close the great divide between OEO (optical-electrical-optical) and OO (optical-optical) switches is to take the hybrid approach and combine electrical and photonic switching in the same box. This is the route Corvis took when it recently unveiled its CorWave Optical Convergence Switch. Others, too, can be expected to take this hybrid approach. But start-up TeraBurst Networks, of Sunnyvale, Calif., believes there's another solution for providing carriers the best of both worlds. At the heart of TeraBurst's OMS (optical management system) platform is its OMO (optical-millimeter wave-optical) switch fabric. TeraBurst is marketing its electrical but transparent switch fabric as one combining the advantages of the OEO and OO approaches, but at a price and performance neither can match.

While the future may lie in an all-photonic network, there are places where OO switches fall short, at least for now. Their disadvantages include limited performance monitoring at best, some loss of light, no benefits from the 3Rs (reshape, retime, regenerate) and of course no wavelength conversion. On the other hand, OEO switches do all this but also have their drawbacks, most notably scalability, a large footprint and higher cost.

OEO switches, says Shantanu Mitra, TeraBurst's product marketing director, make carriers pay a penalty with the number of ports if they decide to upgrade the network. "If you've got a 256 x 256 OC-48 switch and upgrade to OC-192, this would drop the port count to 64 x 64. This would not be the case with OO switches, because if you upgrade an individual line you still consume one channel," Mitra said. Just as the path to a higher port count is easier with OO switches than OEO switches, so is it with the TeraBurst OMO-based switch, according to the company.

TeraBurst describes its switch fabric as electrical analog, in contrast to OEO switches which have essentially a digital fabric, making the latter neither bitrate nor protocol independent. "Analog millimeter wave technology has matured in other industries, such as wireless communications. It has well-understood processes--people understand its failure mechanism--so we didn't have to reinvent any exotic, fancy manufacturing processes to build our switch fabric," Mitra says.

Thus while TeraBurst's technology offers all the advantage of OEO switches, including performance monitoring and wavelength conversion, as a transparent switch fabric it is unlike an OEO fabric, which makes it at once bit-rate and protocol independent. This means it is capable of supporting any line rate (although currently OC-3 to OC-192 is supported) and any networking protocol such as SONET, Gigabit Ethernet or Fibre Channel. "The switch fabric and the backplane is already OC-768-ready. So when the optics are available and viable for 40 Gbps, the only thing that will change to support that are the line cards. This lets carriers do an in-service upgrade while depreciating the base system over a longer time period," Mitra says.

If there is one hit against TeraBurst, it is that unlike many OEO switches, the TeraBurst switch has no subwavelength level grooming capability. According to an official of a carrier that is building a global fiber network, STS-1 grooming was a key reason for it selecting the Ciena CoreDirector switch. The official said that, while his company currently does not have plans to test the TeraBurst product, it does have on-going discussions with TeraBurst. "We want a switch that scales and that maps to the transport bandwidth, which could be anything from OC-3 to OC-192," the official said, noting that the TeraBurst technology does combine the many advantages of OO and OEO switches.

TeraBurst argues that it's gotten mixed messages on this issue from the industry. "The voice people say they want grooming, but they're unclear about how much of a requirement they have for grooming. On the data side, however, there is a clear push to stay with a whole wavelength. When you go to a higher number of channels, you can afford to be a little less efficient with each wavelength," Mitra said. TeraBurst, which has a partnership and has done an interoperability test with Fremont, Calif, start-up Atoga Systems, believes that grooming is best left to a device like Atoga's Optical Application Router, which is a metro access product that combines IP routing, SONET muxing and optical transport.

The TeraBurst OMS will come in two versions. The smaller OMS2100, intended for metro/regional applications, provides 200 Gbps of switching capacity (scalable to 800 Gbps) in one-third of a rack. The OMS2200 is targeted for backbone applications and provides 640 Gbps (scalable to 2.56 Tbps) in one-half of a rack. The company also claims that its platform has far lower power requirements than either OEO or OO switches. According to TeraBurst, the 640-Gbps version requires 2600 watts. An OO switch offering the same capacity would need 3800 watts, an OEO switch 12,000 watts, it claims. The switch is expected to be in customer lab trials soon, with general availability scheduled for year-end.