How fast can data travel over Ethernet? If you answered 10 Gigabit per second, then you would be off by about 90 gigabits per second. Infinera, a San Jose, Calif.-based start-up, along with University of California, Santa Cruz, Internet2 and Level3 Communications, today demonstrated a 100 gigabit/second Ethernet connection that could carry data over a 4000 kilometer fiber network. The trial took place at the Super Computing Show in Tampa, Florida.
The experimental system was set up between Tampa, Florida and Houston, Texas, and back again. A 100 GbE signal was spliced into ten 10 Gb/s streams using an Infinera-proposed specification for 100GbE across multiple links. The splicing of the signal is based on a packet-reordering algorithm developed at the University of California at Santa Cruz. This algorithm preserves packet order even as individual flows are striped across multiple wavelengths.
A single Xilinx FPGA implements this packet numbering scheme and electrically transmits all ten signals to ten of Finisar’s 10 Gb/s XFP optical transceivers which in turn convert the signals to optics. These signals are then transmitted to an Infinera DTN DWDM system.
These packets which have a special sequence numbering are then reassembled by the receiver. In short, Infinera has bonded 10 parallel 10 Gb/s channels into one logical flow while maintaining packet ordering at the receiver. (Services that combine multiple wavelengths to offer a single service are referred to as super-lambda services.)
The trial shows that seriously high speed services can be offered over existing 10 Gb/s transport networks. “Gigabit Ethernet will be a critical technology to accommodate bandwidth growth, and this demonstration shows that we have the capability to implement this as a super-lambda service over today’s networks,” said Infinera co-founder and CTO Drew Perkins.
It was about 14 months ago we wrote about the 10 GB/s network that connected the University of California, San Diego and NASA Goddard Space Flight Center over a dedicated optical path. We chatted with Perkins earlier today and he said that the trial today shows that you can build scalable systems that can achieve higher speeds. “The way bandwidth demand is growing right now, we need 100 Gb/second now,” he says. “The network is going to keep growing.”
The IEEE has set up a Higher Speed Study Group (HSSG) group and is looking for specifications for higher speed Ethernet. The 10 GbE took about five years to become a standard, but the working group might have to hurry this time around. Proliferation of higher speed consumer connections, and emergence of video over IP will make 100 GbE a must have for most carriers.
While the 100 GbE technology is unlikely to show up in your neighborhood anytime soon, the trial is an important step in addressing the ever growing demand for bandwidth. The technologies like the one demonstrated by infinera and its partners will first show up in long haul networks. The largest IP backbones are currently using multiple 10 Gb/s links between core sites.
The online video explosion is going to tax the current fiber and network infrastructure, and will fill up the networks rather quickly, officials at Level 3 explained, when we met with them last week at the Web 2.0 Summit in San Francisco.
28 thoughts on “Ethernet zooms to 100 Gigabit speeds”
US seems to lag in Broadband speeds compared to some European and Korea, is the industry deliberately doing this for profits or is it just the way it is? I hope if they put their heads together and increase the speed it will be beneficial for everyone, we will see more progress.
Its because the US is just so much bigger than any individual country in the EU and Korea. Broadband still hasn’t reach some areas of the US including most rural Texas.
Still using 10 wavelenghts so the key here is the smart reordering algorithm.
IMHO the real breakthrough will be with 100GE over a single wavelenght.
This is not 100G… it is nearly impossible to reach 100% utilization on a multi-link transport path without adding an unacceptable amount of latency.
It’s a vanity project. Big packets would break this configuration.
vanity project…cute. It’s progress, folks. Laboratory settings, but progress. All steps in the right direction, in an effort to keep up.
Bummer, I just finished re-wiring my house for gigE so it looks like I have another project for this weekend. Seriously though, Andrew is right, it’s just a vanity project but pretty cool nonetheless.
hey, does this mean that sun is gonna bring project corona back to life and start selling us all cool smart cards that we can carry our ‘state’ around on to do stuff on dumb terminals calling into super servers via high speed connections? or will elgoog get there first 😉
…at least I would not call it 100Gbe
I’m a big fan of Finisar and have a discussion forum just for Finisar here http://investment.suite101.com/discussion.cfm/179
Very cool! Thanks for reporting Om.
I would suggest it is as much progress as it is vanity…what did Om say 14 months since reporting on 10 GB/s networks…to me that is telling. I suppose we will see.
I remember having a discussion with a Vivato engineer about 2 years ago at a customer conference, and he couldn’t help but laugh when I suggested home Internet speeds of 30+Mbps in 3 or 4 years, based on Metcalfe’s quote of a history of (an average) 50% increase a year in home connection speeds (remember this is pre-Fios/Lightspeed). Of course Vivato is out of business now and Internet speeds are quickly approaching that rate, and some at 30+ already. Yea, it’s just cool right now, but obviously this where things start, and apparently at ever quicker rates.
While Infinera’s FPGA inverse-mux is cool, and Drew is one of the smartest guys around, the development is essentially a work-around for the inadequacy of the US telecom infrastructure.
NTT has demonstrated 111Gbps per lambda in a field deployment over 160Km.
Rather than heaping praise on a company that has a technology hack to circumvent the limitations of the carrier, why aren’t we heaping scorn on the carrier for being tardy in deploying cutting edge technology.
[Each lambda was] modulated at 111 Gbps using the CSRZ-DQPSK
160-km transmission was successfully achieved by amplifying these signals in newly developed optical amplifiers.
NTT demonstrated in this experiment, for the first time, that it is possible to transmit 100 Gbps signal with forward error correction2 bytes and management overhead bytes of the OTN3 frame over long distances allowing the construction of large capacity optical networks that offer 10 Tbps or more.
I believe that entrenched incumbent service providers deliberately slow the pace higher bandwidth consumer offerings. They certainly fall well behind the pace of Moore’s Law. Now, with VoIP and IPTV, and other potentially vastly profitable content on demand services that require more bandwidth in the home, we are starting to see higher speed DSL offerings, Fiber (FIOS) and other services.
Those who own the networks do not innovate to deliver greater value to the consumer. Those who own the networks innovate, begrudgingly, in order to facilitate new revenue streams.
Great. We go from discussing an interesting tech demonstration to pissing on American infrastructure.
Those who own the networks can do whatever the f*ck they want with them. That’s what ownership is. If you don’t like it, move to North Korea.
You guys wondering why the US is so behind when compared to Europe and Asia should take a look at some of the reasons.
This is a great PBS doc about Net neutrality and the fiber network the telcos promised to build us in the 90s but didn’t:
Forgot to add this Google video link:
Well, I can’t think of any who have done what they want with their networks, and are actually still around, at least at their previous stature, or aren’t teetering from debt, but that’s right, they can do exactly what they want with them, at their own risk. I might not like what they have done, but I can’t see anything wrong with it. Who among them are NOT in business to make money?
@chris writes “US seems to lag in Broadband speeds compared to some European and Korea, is the industry deliberately doing this for profits or is it just the way it is?”
One of the main problems is that the US is very large and sparsley populated. Something like 90% of South Koreans (I assume you weren’t including North Korea. They just got hamburgers there, I’m certain they don’t have FTTP yet) are found in three relatively dense metro areas. Not only are there no Iowas to deal with, there’s not even a Connecticut or New Jersey.
So no, there’s not a telecoms industry plot to somehow increase profits by NOT selling customers the services they want. There’s geography and demographics.
Great to see all the comments here on the Infinera 100GbE demo. I must point out–this is not a vanity project, but a serious effort to develop and propose a solution that works for higher Ethernet speeds for the entire industry. We do believe the issue of very large packets can be dealt with effectively. See further details from my colleague Ted Sprague at http://www.infinerablog.com.
Note, people should really follow the link to the Infinera site. I initially assumed this was link aggregation over 10x 10GE signals (see comment above) and I was wrong. This is a real 100GE solution, albeit proprietary, but with efforts underway to standardize it.
What is VERY interesting is the high integration inherent in Infinera’s approach gives them a distinct advantage in this type of solution.
Lo que ta en chino no es verdad
More progressive noted today in 100Gig on LightReading. Like I said…every effort made, such as Infinera’s, is progress. So much for the vanity project/”big packets” theory, whatever that meant.
I was wondering what some of the more mature members here do about dating. It seems much harder for older singles to find a mate, so I might be turning to online dating for older singles.
any suggestions? thanks.