100G Ethernet

Ciena, Infinera Push 100-GigE

Companies itching to be identified with next-generation transport -- 40- and 100-Gbit/s stuff -- are stepping up their efforts in research circles, as reflected in recent announcements from Ciena Corp. (NYSE: CIEN) and Infinera Corp. (Nasdaq: INFN). (See Internet2, Ciena Team.)

In particular, the supercomputer community is interested in higher speeds, not just for the interconnection between computers, but as the link between sites.

That's one of the targets for Boston South Network (Bossnet), a long-reach optical network running between Boston and Washington, D.C.

"What they really see supercomputers interconnecting with is 100-Gbit/s Ethernet, so they want to put that on the transmission network," says Ciena's senior VP of strategic planning, Thomas Mock.

Bossnet is considering the move to 100-Gbit/s Ethernet, but, for now at least, the plan is to upgrade the network to 40 Gbit/s early in 2007, a step up from the 10-Gbit/s Ethernet it's using now. Ciena announced this week it's been chosen for that upgrade. (See Ciena Upgrades Bossnet.)

That interim step is all down to experience, "because 40 Gbit/s is something we, and others, have done a lot of research on," Mock says.

Meanwhile, 100-Gbit/s Ethernet is inching towards reality. Recently, the Institute of Electrical and Electronics Engineers Inc. (IEEE) 802.3 Higher Speed Study Group (HSSG) officially voted to make 100-Gbit/s Ethernet the subject of its standards work. (See 100-Gig Ethernet Gets Official Nod.)

Lucent Technologies Inc. (NYSE: LU) has put up a few papers on 100-Gbit/s transmission this year, though with work that's primarily at the optical layer and conducted at Bell Labs . (See Lucent Stretches 100-GigE.)

And earlier this month, in conjunction with the recent SC06 supercomputing conference, a group of vendors and carriers demonstrated 100-Gbit/s Ethernet traversing the live Level 3 Communications Inc. (NYSE: LVLT) network between Houston and Tampa, Fla. (See Group Demos Technology.)

Infinera is calling it the first real test of 100-Gbit/s Ethernet on a live network.

Other folks involved with that demo include Finisar Corp. (Nasdaq: FNSR), Internet2 , and the University of California, Santa Cruz.

(Ironic that the banana slugs had a hand, so to speak, in a 100-Gbit/s Ethernet test.)

For the demo, the 100-Gbit/s signal was split into 10 lines of 10 Gbit/s each, the kind of division that's likely to represent 100-Gbit/s Ethernet's first incarnations. One motivation behind the demo was to show that 100-Gbit/s Ethernet is possible on the current 10-Gbit/s infrastructure, says Drew Perkins, Infinera's chief technology officer.

"We see it as being a big part of the future," says Perkins, adding that Infinera expects Sonet to be "displaced over time with an Ethernet-based transport network."

In terms of technology, the 10x10-Gbit/s configuration is missing one major piece -- "having a chip that sees all 100 Gbit/s," Perkins says. Commercially available media access controller (MAC) chips don't go higher than 40 Gbit/s. Or so it seemed. "Now it turns out the very latest FPGAs are, in fact, capable of [seeing all 100 Gbit/s]," he adds, noting that the group used Xilinx Inc. (Nasdaq: XLNX) chips for the demos.

As for serial 100-Gbit/s Ethernet, Perkins isn't so sanguine. "We don't believe that is the most cost-effective approach," at least not within the next five years, he says.

— Craig Matsumoto, Senior Editor, Light Reading

Page 1 / 3   >   >>
ruready 12/5/2012 | 3:33:45 AM
re: Ciena, Infinera Push 100-GigE The technology sounds good, especially for the system vendors. However, the component vendors need volume to make money. With the 2007 forecast for 40G at ~1K units, nobody makes money.
materialgirl 12/5/2012 | 3:33:42 AM
re: Ciena, Infinera Push 100-GigE Volume is one thing. Money is another. DOD is big into supercomputers for nuclear weapons testing. They have budget to spend tens of millions of dollars on any number of projects. They want that 100Gbps. That is a start that has spawned more than one new market.
SolitonWave 12/5/2012 | 3:33:40 AM
re: Ciena, Infinera Push 100-GigE 10x10G is not 100G. Two basic differences: - Wavelength usage; -need for link aggregation or inverse multiplexing.
The evolution step for 40G is 100G (serial). That's the nominal bitrate per wavelength. Nobody would say that a vendor demonstrating 4x10G today would be "pushing 40G technology". So claming that by doing 10x10 "Infinera is pushing 100G" is not technically correct.
Sure, they integrate all the stuff in a chip but they still need 10 lambdas and some kind of link aggregation, so we are talking of two basically different technologies that have a set of different problems to solve.
stephencooke 12/5/2012 | 3:33:39 AM
re: Ciena, Infinera Push 100-GigE "But seriously LR, the "100G" title really is misplaced, unless you're going to start labeling other WDM technology by its aggregate bandwidth."

In DSL this technology is called "bonding" (see G.998.x). It might be better described as "logical concatenation". Making more efficient use of existing technology seems to be the right way to go for these "one of" projects. Applying logical concatenation to optics is a great idea and a decent story. To get the "buzz" in today's press though, you have to have big bandwidth numbers.

Stevery 12/5/2012 | 3:33:39 AM
re: Ciena, Infinera Push 100-GigE So you mean Ciena and Infinera are not being fully up-front? Are they engaging in spin? I'm shocked!


But seriously LR, the "100G" title really is misplaced, unless you're going to start labeling other WDM technology by its aggregate bandwidth.
litereading 12/5/2012 | 3:33:39 AM
re: Ciena, Infinera Push 100-GigE So you mean Ciena and Infinera are not being fully up-front? Are they engaging in spin? I'm shocked!
litereading 12/5/2012 | 3:33:39 AM
re: Ciena, Infinera Push 100-GigE MaterialGirl,

Exactly! Before you know it they will be installing supercomputers into the attic or basement of every new home. Won't that drive volume deployment of FTTx at 100G levels....

Pete Baldwin 12/5/2012 | 3:33:38 AM
re: Ciena, Infinera Push 100-GigE I seem to recall 4x10G being called "40 Gig" back in the day, although I'll admit my memory could be faulty. And I'm told there was once a Frame Relay 4x2.5G effort that presumably would have been called "10 Gig."

You're welcome to take the "100G" moniker with a grain of salt, but I do think non-serial 100GE is an effort worth following.
materialgirl 12/5/2012 | 3:33:38 AM
re: Ciena, Infinera Push 100-GigE Dear litereading:

Don't laugh. The INQ site just reviewed a PC with TFLOPS power running at 500W. It does all sorts of real-time rendering. They called it a desktop supercomputer. Read it for yourself.

Besides, DOD funded ARPA. That started the Internet if memory serves correctly. They can fund 100-GigE too, if it serves their interests. Apparently it does, given their current budget spends on supers for these nuclear simulations. We have to keep up with N. Korea, after all.
gadfly 12/5/2012 | 3:33:37 AM
re: Ciena, Infinera Push 100-GigE At the heart of the technical challenges is that Infinerap apparently cannot get direct modulation of their integrated lasers to 100Gb/s
each with good channel separation and at cost effective levels.
Nice that the existing fiber in the SC06 trial was able to handle it, although Bell Labs and Corning would rather not hear that...
Page 1 / 3   >   >>