100G Ethernet

100-Gig Ethernet Gets Official Nod

Not that it's a huge surprise to anyone, but the next standard speed for Ethernet will be 100 Gbit/s.

That's the vote from the Institute of Electrical and Electronics Engineers Inc. (IEEE) 802.3 Higher Speed Study Group (HSSG), taken during last week's meetings in Dallas.

It's a formality, but a necessary one, as a few other options had been mentioned since the HSSG started. (See 100-Gig Ethernet Takes First Step.) "Now we can move foward" with the work of trying to build a standard, says John D'Ambrosia, the Force10 Networks Inc. representative who's chairing the HSSG effort.

Many discussions pitted the 100-Gbit/s option against the possibility of 40-Gbit/s Ethernet, a speed in step with the OC768 of Sonet/SDH. But that fight didn't turn out to be serious; rather, a suggestion of 120 Gbit/s was the only realistic competitor, according to D'Ambrosia.

"There was a lot of consensus around the fact that 40 Gbit/s wasn't the speed to choose, and there was the same amount of consensus around 80 Gbit/s," he says.

Quite a bit of support has been behind the idea of 100-Gbit/s Ethernet, as shown at a recent Optoelectronics Industry Development Association (OIDA) seminar on the topic. (See 100-GigE Takes Shape.) Moreover, data centers already appear likely to adopt 100-Gbit/s Ethernet as a means of connecting machines in high-density environments.

"It's unlikely in the longer term that telecom networks wouldn't gravitate to what's being used in the data networks," says Tom Mock, senior vice president of strategic planning for Ciena Corp. (NYSE: CIEN).

Of course, 100 Gbit/s got all the good publicity along the way. It's a nice round number, and companies like Lucent Technologies Inc. (NYSE: LU) have been seeding the news wires with tales of 100-Gbit/s research. (See Lucent Stretches 100-GigE.) Representatives of the numbers 40, 80, 120, and 160 are expected to demand a recount.

— Craig Matsumoto, Senior Editor, Light Reading

Page 1 / 4   >   >>
mugwhump 12/5/2012 | 3:34:13 AM
re: 100-Gig Ethernet Gets Official Nod It will be interesting to see who leads the charge in getting to 100GE and is first to market. There are some vendors who have bet the farm on 40, some who have been working and hoping for 100 and I suspect a bunch that have been sitting on the sidelines.

Any bets?
Pete Baldwin 12/5/2012 | 3:34:12 AM
re: 100-Gig Ethernet Gets Official Nod Yes, this is where the real fun starts, and the "pre-standard" implementations too.

Let's see -- off the top of my head, I remember Extreme, Foundry, and Force10 all talking about being 100 gig ready. I suppose Infinera is, technically, but they're not exactly an Ethernet play, so they don't count.

Lucent's been doing a lot of 100GE lab work, of course.

I guess the more interesting question is who gets set back by this. Although, given the three-year schedule for an IEEE standard, you'd think systems players would have time to recover.
krbabu 12/5/2012 | 3:34:12 AM
re: 100-Gig Ethernet Gets Official Nod As part of this standardization, is it known that the 100 Gbit/s Ethernet will be transportable over the wide area as well, and not just in the data center? If so, there is a chance that SONET will become merely an access mechanism over such a 100 Gbit/s Ethernet core network.
tsat 12/5/2012 | 3:34:11 AM
re: 100-Gig Ethernet Gets Official Nod
How will they do 100GE on a fiber? One wavelenght? 8? 10?

Pete Baldwin 12/5/2012 | 3:34:11 AM
re: 100-Gig Ethernet Gets Official Nod Ha! Now I've gone and made Infinera mad. Sorry about that guys -- I was only commenting on the fact that you're not a switch vendor (unless there's something you're not telling us...)

But I should point out the 100-gig demo from SC06, which involved Infinera, Level 3, & others:

vmg00 12/5/2012 | 3:34:10 AM
re: 100-Gig Ethernet Gets Official Nod krbabu,
If that is the case SONET will be wiped out as in access SONET is being replaced by Ethernet slowly.
if in Wide area also PBT or TMPLS beats SONET than SONEt is dead i think.
Pete Baldwin 12/5/2012 | 3:34:10 AM
re: 100-Gig Ethernet Gets Official Nod Right now, I think it's a case of picking your favorite factors of 100: 4x25, 5x20, 19x5.26 (OK, maybe not THAT one)...

But yes, that's a bigger question that's going to take some real work.

Of course, the ultimate goal is going to be some kind of 100GE serial, right?
vmg00 12/5/2012 | 3:34:10 AM
re: 100-Gig Ethernet Gets Official Nod tsat,u dont need 8 or 10 wavelength ?
What is the problem in carrying 100G on one wavelength, only thing in DWDM environment u can carry less wavelength now.
I think NTT is their latest 14Tbps experiment used 100G
tsat 12/5/2012 | 3:34:09 AM
re: 100-Gig Ethernet Gets Official Nod A single wavelength might be too expensive. Might be cheaper to DWDM them.

CoolLightGeek 12/5/2012 | 3:34:09 AM
re: 100-Gig Ethernet Gets Official Nod The death of SONET has been prematurely predicted for a long while.

Many of today's Ethernet services are provided over SONET backbone networks.

In engineering terms, SONET is the real pseudo-wire. The data transmission characteristics (throughput and delay) it provides per user channel are rock solid and won't "weaken" as carriers load on services or add more nodes to a network. In addition, the transmission characteristics for SONET/SDH are universally the same across the world.
A network engineer knows the channel characteristics will stay rock solid independent of how many carriers the channel has to pass through. There is NO concern about engineering for how many switches the SONET signal has to go through. Performance monitoring is typically a matter counting zero errors unless a laser goes bad or a "once a year" 50 ms protection switch.

Then consider complexity of characterizing a statmux channel that passes through two or more service providers (each provider may have any number of packet switches), each carrier choosing there own stat mux optimizations. Then think about how you as an end user might actually test that all the carriers involved will actually be able to deliver your minimum guarenteed data throughput through all the permutations of network stress rhat carrier might get.

Then think about the service QOS definitions in tariff service of Ethernet as it compares those of SONET.

Ethernet based tariffs are great within a service provider network. They also are relatively new tariffs as compared to SONET and thus the service providers have had more incentive to price them lower than rate reductions on existing SONET tariffs. But handing Ethernet between carriers will take some time to work out the kinks.

Real network traffic patterns are not flat: instead, they tend to hub to and from key service points. On their way to and from the hub points, packet streams are packed into a data stream that has a distance to travel. Once packed, it is a disadvantage to continue to pass them through packet switches on their way across the network.

Getting a packed channel through offices on the way to a hub typically involves making a choice of sending it through a packet switch, a SONET switch or ADM, or a WDM switch or OADM. The choice comes down to how much traffic is involved. If there is enough traffic, a WDM or SONET solution to compliment the packet switching is a better networked option than putting huge packet switches everywhere.

Clearly, Ethernet is displacing SONET as the world becomes IP centric but it will still be significant time before the true multi-vendor multi-carrier packet networks can provide the ubiquitous "gold standard" of network transport service that SONET has provided for at least the last 15 years.

SONET backbone transports- still a very large market.

One might consider SONET's appeal as the FedEx of optical transport technologies:

When you absolutely positively have to get it there as fast as possible, use SONET.

Page 1 / 4   >   >>
Sign In