Slashdot Mirror
World's Fastest Internet Transfer Rate?
vrioux asks: "While browsing through available Internet service offerings in my area, I became puzzled by the large amount of different speed records that seem to have been achieved in the past months. Some say that 5.44Gbits was the fastest ever achieved, while others seem to think 923Mbits is still in the race. Is there anybody who really knows who holds the Internet transfer world speed record?"
... a 747/station wagon/fed ex truck or similar full of $mediaofchoice
Don't blame me, I voted for Kodos
are they measuring. I mean its one thing to have a really fast transfer rate from computer to computer if they are right next to each other, its another thing to go half way around the world.
30% Troll, 50% Underrated, 10% Interesting
Score:5, Troll
last i heard, a truck full of backup tapes was still faster than the internet.
Seems pretty pointless... anyone can come up with all sort of result base on their method of measuring. eg which protocol are you using? (eg http, ftp?) where are your source and destination hosts? time and duration of your test, etc.
Codeala - Just another mindless drone
On a side note, your standard fiber OC-192 is 10-Gbit/s. And your OC-48 is 2.45Gbit/s. While your OC-192s are definitly not common, its not unheard of for service providers with big ONS SONET networks to have them.
"Is there anybody who really knows who holds the Internet transfer world speed record?"
/*I hope the mods are in good humor today*/
No, but I have a compelling essay about who was the best Star Trek captain. I have some time tomorrow night if you'd like me to read it to you.
"Derp de derp."
An OC-192 SONET circuit is (IIRC) 9.6 Gbps and are used in lots of places.
My guess is that they were referring to serial transmittion (i.e. one bit at a time - a "truck with tapes" wouldn't qualify) specifically across the atlantic, using existing circuits. Otherwise, those numbers don't sound like much of a record.
It's entirely possible that these records are beat on routine bases by telco's/ISP's, perhaps they just don't necessarily care to disclose it.
Surprisingly, at least two people seem to care -- vrioux and Cliff. I'm surprised.
Slow news day, eh?
Somebody get that guy an ambulance!
Roughly 100 years ago, the world record for powered airplane speed was probably under 60 mph ... I imagine it's somewhat faster now.
(Of course, `Internet Speed Record' ... who really cares? I'll care when I can have this transfer rate to my house for $60/month :)
...the Internet2 Land Speed Record, which is included in the the Guinness Book of World Records. The actual marks are bandwidth x distance
Never underestimate the bandwidth of a station wagon full of hard drives.
Tat-14
"It has a dual route, transatlantic capacity of 640 Gbits on 2 service fiber pairs backed up by 2 protection fiber pairs. This configuration provides a capability of transporting 4,096 STM-1's or approximately 9,700,000 circuits across the ocean."
It more depends what is being counted though. Your never going to get all of this fiber at once. If you count from a local isp the speed record or from the backbone or from a rual isp it all depends.
Sounds like an appropriate metric to me.
Any numbers you read in any of the above articles should be multiplied by the distance between endpoints across all router links, and re-ranked. The parent post has a good example.
THIS THING CAN TURN ON A DIME, MACROSSZERO STYLE ALSO FUCK BETA, ~NYORON
You can put two desktop machines with good frontside busses side by side, and use etherchannel using the new 10gbit ethernet chips by Intel, and you theoretically have 20gbit. I'm pretty sure upper-end Juniper and cisco ATM switches can do it much better using loadbalancing of interfaces.
So do two systems side by side and connected to the net qualify as 'across the Internet', or is the rule to use only standard consumer ISPs?
"Give orange me give eat orange me eat orange give me eat orange give me you." -Nim Chimpsky
These mystery articles talking up "new internet speed records" are not talking about "mega-fast line capacity". Anyone can engineer a nice fat bundle of lines to up the ante - it just costs Money (specifically, LOTS OF).
What these people are going on about is the real-world actual measured data-throughput between a single pair of computers across "the internet" (usually across a significant chunk of the world).
Anyone who's tried to use any TCP based protocol (eg FTP, HTP) to send mind-warpingly large chunks of data understand the throughput limitations imposed on said communication as a direct result of large RTT latency.
I'm not sure about these two instances specifically (articles give essentially ZERO technical details) but many similar researches are an attempt to derive new (bigger, better, faster, and more eco-friendly) protocols which avoid the limitations of TCP.
Visit CryptoGnome in his home.
It was obviously Picard.
You can't overcome latency. There's a fundamental limit. I mean, satellite communication links have latencies of 1 second because of the speed of light, and you can't "improve" that by much. The latency is a side effect of the distance. Which means that you actually trade-off effective bandwidth when you increase distance because sustaining throughput with error correction gets harder.
So a doubling of distance should get you bonus points! You can overcome system latency with fancy hardware, but you can't fundamentally improve distance effects. So as a pure measure of system performance, I think distance is key to knowing the true prowess of said link.
THIS THING CAN TURN ON A DIME, MACROSSZERO STYLE ALSO FUCK BETA, ~NYORON
This year's Supercomputing 2003 Bandwidth Challenge netted some cool results in this area including 23+gbps inside the US, 7.56gbps between the US and Japan, and 8.96gbps throughput to a remote network filesystem using GPFS. There are even some pretty graphs. My former co-workers at SDSC were involved in a lot of this work.
Most call it the "Internet land-speed record"
More info at http://lsr.internet2.edu/
Also, the fastest router from juniper networks, to my knowledge, is the m160, capable of forwarding 160 GB/s. But, that is spread out over several interfaces.
-n
rather than funny? wacky mods.
01100111 01100101 01110100 00100000 01101111 01110101 01110100 00100000 01101101 01101111 01110010 01100101 00101110
Lookup the FAST (FASTTCP) project at Caltech. The problem is not the physical medium but the communication protocols. They have been working on a TCP variant that manages to fully utilise link capacity (which current TCP cannot do).
All the info, as well as links to TONS of popular press is there. If memory serves, current record is well into the gigabytes.
I have a record setting Terabyte in under an hour with the Pigeon RFC Protocol. RFC 2549
John Walsh once found me while looking for some other kid. He was not amused.
That's nothing, my home computer can trasmit 40 jiggaquads worth of data over gold subspace channel, to the nearest starbase.
If I multiply the speed by distance, that's. Arggh! your puny hu-man Internet2 is like using constupated pigeons!
Those who can, do. Those who can't, consult.