Lucent Stretches 100-GigE

Bell Labs , which has made a hobby out of 100-Gbit/s Ethernet lately, has presented lab results that claim serial 100-Gbit/s transmission across a 2,000km distance.

Bell Labs parent Lucent Technologies Inc. (NYSE: LU) divulged the results earlier today at the European Conference and Exhibition on Optical Communication (ECOC) in a paper titled, "2,000-km WDM Transmission of 10 x 107-Gbit/s RZ-DQPSK." Not that catchy, but maybe you can dance to it. (See Lucent Touts 100 Gbit/s.)

What's important is that Bell Labs ran its experiment using ordinary 40-Gbit/s components in a reasonably real-world setup. The 2,000km span was adorned with Raman amplifiers every 80 to 100 km, just like in real life, and most of the parts used are available commercially.

The experiment follows Lucent's earlier accomplishment of just transmitting 100-Gbit/s Ethernet at all. That test, first noted last September, wasn't done in quite real-world conditions, using tricks like optical time-divsion demultiplexing, which is "a little bit cheating," says Bell Labs director Martin Zirngibl.

The 2,000km test, by contrast, used commercially available chips for its multiplexing and demultiplexing. The DQPSK modulation let Bell Labs produce 100 Gbit/s of data on signals running at just 50 GHz -- the latter being a speed that current 40-Gbit/s chips can handle.

There's still no guarantee that 100-Gbit/s Ethernet will be a standard. That's up to the Institute of Electrical and Electronics Engineers Inc. (IEEE) Higher Speed Study Group, which will debate the issue in the coming months. (See 100-GigE Takes Shape.)

Still, it's crucial to the 100-Gbit/s cause to show that the technology can work.

"The transmission part is just one piece of 100-Gig, but it's important because if you can't cost-effectively transmit serial 100-Gbit/s, then there won't be any 100-Gbit/s systems," Zirngibl says.

The most immediate interest in 100-Gbit/s Ethernet is coming not from the long-haul sector, but from data centers, where heavy bandwidth loads are creating a need to switch Ethernet in larger chunks. That presents a bit of PR difficulty whenever the outside world clamors to hear about a 100-Gbit/s killer app. "The problem we found is that the killer app is aggregation," said John D'Ambrosia, components technology scientist for Force10 Networks Inc. , at a recent Optoelectronics Industry Development Association (OIDA) seminar on 100-Gbit/s Ethernet. (See 100-GigE Takes Shape.)

Zirngibl says the next step for 100-Gbit/s Ethernet would be to look inward, determining how to deal with high-speed signals on the backplane. Serial 100-Gbit/s Ethernet backplanes don't seem likely, so alternatives such as four lanes of 25 Gbit/s are being considered.

— Craig Matsumoto, Senior Editor, Light Reading