Tokyo team busts the Ethernet speed record
Fastest IPv6 network in the world
A team led by Toyko University researchers has set a new "land speed record" for IPv6 Ethernet, transferring data over a distance of 30,000km at an application rate of 9.08Gbit/s.
The team leader, Dr Kei Hiraki, claimed this will not be beaten for the 10Gig generation because the contest organiser, the Internet2 consortium, demands that each new record shows a 10 per cent improvement over its predecessor.
The team actually broke the record twice - first sending 7.67Gbit/s using standard protocols, and then two days later sending 9.08Gbit/s using modified protocols - and they did so in both the single and multi-stream classes. The record-breaking run took place last December but has only just been reported, at a recent Internet2 members meeting in the US.
"These records are final for the 10Gbit/s era because they represent more than 99 per cent of the upper limit of network capacity," said Dr Hiraki.
He added that the test crossed six international IP networks, thanks to the team also including the WIDE Project, NTT Communications, JGN2, SURFnet, CANARIE, Pacific Northwest Gigapop, and other institutions. Data travelled from Tokyo to Chicago, on to Amsterdam, then Seattle, and finally back to Tokyo.
According to Chelsio, which developed the TCP-accelerating Ethernet adapters used by the team, the new record shows that IPv6 network devices are able to provide the same performance as IPv4 or better.
The speed record for IPv4 currently stands at 8.8Gbit/s, so could, in theory, still be surpassed. However, the next target for wire speed record junkies is rather more likely to be the upcoming 100Gbit/s spec for Ethernet, which is under development. ®
Baam! Richard said it
But, a small point is that the most of the world uses SDH, not SONET - so the channel would be a STM-64. But then I think the rates are the same anyway, making this a rather useless contribution.
Educate yourself a bit...
Obviously Goldie needs to pull his head out of a dark place for a while and to educate himself about how the Internet works.
Dr. Hiraki's team transferred data over a distance of 30,000Km.
Last time I've checked, there's no "Ethernet" standard that allows transmission over such a distance.
The record established is an *Internet* Land Speed Record, not an *Ethernet* Speed Record. In the real world, a 30,000Km data link spanning several continents will have to use OC192 optical channels supplied by a telecom carrier, unless you're willing to lay your own optical fiber spanning several continents as well as the associated signal regenerators / repeaters.
OC192's data rate is 9953.28 Mbit/s (of which 9621.504 Mbit/s is the raw payload and 331.776 Mbit/s is taken up by overhead). From this 9621.504 Mbit/s raw payload, you must then substract the IP and TCP encapsulation overhead to arrive at the data rate that is really available to an application.
The new application-level record of 9.08Gb/s is thus close to a practical maximum.
Do some math yourself
Obviously Dr. Hiraki needs to leave the university for a while and to visit some school - he needs few lessons in reading and mathematics:
a. there are many many standards for 10Gbit Ethernet physical layer (OSI L1) but the most popular one does support two bitrates - 9.953 and ... 10.3 Gbit/s;
b. 10% over the achieved 9.08 makes 9.988 Gbit/s which indeed cannot be exceeded on a 9.953 but should not be a big issue for a 10.3 Gbit network.
If one does not trust the biggest team blog on Earth (http://en.wikipedia.org/wiki/10_gigabit_Ethernet#LAN_PHY), just looking at some vendor specifications (http://www.digchip.com/datasheets/parts/datasheet/368/PM5390.php / http://www.pmc-sierra.com/products/details/pm5390/) should prove the existence of the ten-plus-epsilon bitrate.