Ready for 100-Gig Ethernet?

NEW YORK — Ethernet Expo — Is it too early to talk about 100-Gbit/s Ethernet? Apparently not, judging from speakers here.

The move to "100 Gigs" is a foregone conclusion, the only question is when, according to most experts.

"Where does it go next?" asked Bob Metcalfe, the inventor of Ethernet, during his keynote address. "The tradition in the Ethernet world is factors of 10. At each stage folks say, we don't need to go to the next level. But we always do. We're going to 100G."

Experts on yesterday's Ethernet Expo panel session on 10-Gbit/s Ethernet concurred. A 100-Gbit/s speed grade seemed to be the popular choice. Moreover, (NYSE: LU) recently demonstrated 100-Gbit/s Ethernet, showing the technology might not be so far off. (See Lucent Tries 100 GigE .)

But wait -- it's not a slam dunk. As 10-Gbit/s Ethernet begins its mainstream run, no succeeding standard is yet in the works. That leaves open the possibility that a 40-Gbit/s Ethernet could be picked as the next speed.

Indeed, Metcalfe points to 40-Gbit/s as an option, given that it's the next speed in the DWDM and the Sonet worlds. In fact, Metcalfe pointed out that he wouldn't mind if 40-Gbit/s came first, if only because as a VC at Polaris Venture Partners, he's funded 40-Gbit/s DWDM startup Mintera Corp.

Just a few years ago, it seemed pragmatic to have Ethernet match Sonet's OC768 speed grade. But in recent years, that fever has died down, and carriers have found they can use multiple 10-Gbit/s lines as a cheap and practical alternative to pioneering 40-Gbit/s.

All told, vendors on yesterday's panel said they'd prefer jumping straight to 100 Gbit/s, particularly given the amount of time it takes to complete an Institute of Electrical and Electronics Engineers Inc. (IEEE) standard.

"Think about the time and cost of pushing these things together," said Stephen Garrison, vice president of marketing for Force10 Networks Inc.

There's a future-proofing argument as well. A backplane ready for 100-Gbit/s Ethernet would last at least one generation longer than a 40-Gbit/s offering. Vendors are already taking that step, with Foundry Networks Inc.'s (Nasdaq: FDRY) latest boxes packing 100-Gbit/s-ready backplanes. "New products coming to market today are already anticipating techs that are not here yet," said Bob Schiff, Foundry's vice president of marketing.

But does anybody need 100 Gbit/s of anything, yet? Actually, yes, and it's not a sci-fi future kind of vision, either. Internet exchanges in Amsterdam and Tokyo already need 100-Gbit/s trunking, said Eric Troyer, senior product manager for (Nasdaq: EQIX).

"They have an aggregate usage, as of yesterday, of 97.8 Gigs across their backbone fabric," Troyer said of the Amsterdam exchange.

Coming back to the here-and-now, panelists noted that 10-Gbit/s Ethernet is proceeding smoothly on its journey to mainstream bliss. Their optimism was hardly a surprise, given their collective vested interest in the technology, but there's evidence to back them up. Prices have fallen dramatically during the past year, and a supporting infrastructure is forming. "Many companies are working on 10-Gbit/s NICs from Intel on down the food chain," said Force10's Garrison.

In fact, 10-Gbit/s NIC pricing "is dropping faster than port pricing," Schiff said, thanks to competition from startups including Chelsio Communications Inc. and Level 5 Networks Inc.

— Craig Matsumoto, Senior Editor, Light Reading

mr zippy 12/5/2012 | 2:57:31 AM
re: Ready for 100-Gig Ethernet? How many users/customers are going to suffer if a 100G trunk fails ? Sure, you could put another one in for redundancy, however that is an awful lot of capacity to have just sitting there doing nothing to cover a failure scenario.

Running both links at less than or up to 50% utilisation to allow for fail over in case one of them breaks isn't any better, although the bean counters might like it better because at least neither of the links is sitting completely idle.

Where is the threshold where the impact of a link failure effects too many customers, and where protecting against that impact becomes too uneconomical ?

Certainly, providing 100Gbps of capacity via 10 x 10Gbps links is complicated, and possibly may even cost more in equipment and administration terms, however it does significantly reduce the number of customers that would be effected by a single link failure, and it means that it can be cheaper to provide redundant capacity to cover a 1 in 10 link failure, assuming that not all those 10Gbps links are not being carried over the same fibre.

Would 100Gbps Ethernet be limited to a niche market, or would it be widely useful ?
tsat 12/5/2012 | 2:57:26 AM
re: Ready for 100-Gig Ethernet?
While I think the technology for a 100G switch blade is not too far out.. putting a 100G blade on a router is much more difficult. Would 100G catch on without line-rate 100G router ports?

OldPOTS 12/5/2012 | 2:57:21 AM
re: Ready for 100-Gig Ethernet? Many Switches and Routers are based on 40 Gig backplanes, assuming it would be the next uplink standard. Doesn't this mean a couple of 40 Gig alternate uplinks serve the needs for a while?

Sign In