Slashdot Mirror
Net Speed Record Smashed
BrianWCarver writes "The BBC is reporting that scientists have set a new internet speed record by transferring 6.7 gigabytes of data (the equivalent of 4 hours of DVD-quality movies) across 10,978 kilometres (6,800 miles), from Sunnyvale in the US to Amsterdam in Holland, in less than one minute. Average speed: more than 923 megabits per second, or more than 3,500 times faster than a typical home broadband connection. The data was sent across the Internet2 network. Stanford Linear Accelerator Center (Slac) Computer Services participated in the record-breaking event. Slac has an interest in such high-speed transfers as they have accumulated the largest known database in the world, which grows at one terabyte per day."
I'd like to know what media they used that could write that much information in 1 minute.
Well, I guess the're not running experiments every day. Otherwise, when will they find the time / cpu power needed to parse all of this ? Are we goingg to see a SLAC-athon@HOME any time soon ?
All generalizations are false, including this one...
1 tb is 1024 * 1024 * 8 megabits (8388608 MB)
8388608 MB / 923 MB/s = 9088 s
9088 s = 2.53 h
Seems fast enough, or am I missing something?
Yes, OC192 is faster, but I've never heard of a single computer being able to push that much data that fast over a single connection.
A few years ago SGI did a test where they leased a piece of cross-country dark fibre for a day and ran GSN over it. That's a single connection-- using ST, not TCP-- from one computer to another computer, RAM to RAM. They pegged over 790 MB/s (that's a big B, as in megabytes per second), and sustained it for hours. And, just to reiterate, this was from one computer to another computer, without any fancy-schmancy multiplexing or anything. This was the ST equivalent of a single FTP transfer.
I can't find any documentation of this test on the web, but I saw it with mine own eyes. One end of the connection was in Herndon, VA, (where I was) and the other was out west someplace.
The SLAC test did 900+ Mb/s over a switched network, which is darned impressive. It's undoubtedly a record for a public switched connection. But don't go thinking it's an absolute land speed record or anything like that.
I write in my journal
Does this mean that the old adage (that driving across the Atlantic in an old beat-up VolksWagen kombi filled with CD's provides greater bandwidth than the internet does), is no longer true?
"Those who don't believe in magic will never find it." - Roald Dahl
I'm still looking for evidence that we're decreasing the typical latency to get important info (like the fact that I just shot my sniper rifle at some counter-terrorist) across the globe.
Imagine, international internet gaming with low latencies all 'round. Sounds like a pipe dream.
-- Mojo Tooth : exploring our world as only an idiot can.
Lesse...
10,978Km / c (speed of light) = (about) 0.0036 s
At least 3.6ms latency. Likley in the 5ms range tho, considering cut-through times and propegation delays.
Soko
"Depression is merely anger without enthusiasm." - Anonymous
I'm wondering if they did this using TCP, and if so, how the test was set up.
Eg is this the peak rate that it was able to sustain for a one minute period once the transfer was already running, or did it take one minute from start to finish. It's an important distinction with TCP because slow start needs several round trips in order to open the window large enough to get max througput over such a high speed, long distance link.
Also how on earth did they handle packet loss? Getting the max throughput out of a high-latency link with just a single TCP connection is not easy.
Seriously though, if anyone here is thinking that any ISP with 3,500 256kbps connections has that bandwidth available, think again. The issue of contention comes to play. Here in the UK, the contention ratio for a typical home DSL connection is 1:50, meaning the ISP allocates one fiftieth of (number_of_connections x connection_speed). In other words, the ISP oversells their capacity 50-fold. Business DSL connections get more bandwidth reserved for them, typically 1:20 here.
You can't shut us down! The Internet is about the free exchange and sale of other people's ideas!
Why do people constantly talk about n-DVD-hours worth of data? Particularly since they are generally referring to DVD-Video, not DVD data.
Comparing the transfer capacity to some number of hours of DVD video material is pointless, since the bitrate is not the same from one title to the next.
For example, 6.7 gigabytes of data is actually only 6.23 gibibytes. A video stream would have to be encoded at around 3.5 mebibits/second to fit four hours of material in 6.23GiB. I wouldn't call that a quality video stream. And that's WITHOUT an audio sub-stream! You're not far away from Super VCD world at this bit rate.
Now, using a more reasonable average bitrate of, say, at least 4.5 mebibits would mean that the 6.23 gibibytes of data would only hold about 3 hours of "DVD-Video quality material".
Which brings us back to my point. Using DVD Video as a measure of data capacity is pointless, since there is no single data rate used for DVD Video.
* As is generally the case, my opinions do not reflect those of my employer.
The question is which government? Last time I looked a) the Internet was international and b) we din't have a world governement.
I am TheRaven on Soylent News
This is probably the first valid Beowulf post in about 2 years...
If you have a master node acting as round-robbin server, you could have hundreds of machines behind it. Each of those, in turn, could be the master node of a large Beowulf cluster.
Or just picture your ISP's core switch. It is transfering the data for thousands of users. That data is being read and written, just not by one computer...
I'd rather you do it wrong, than for me to have to do it at all.
TCP in particular - the standard window sizes have a maximum of 64KB, which means that a single TCP session can't run a 10,000km pipe very fast - speed-of-light-in-fiber latencies are about 60ms one way, so do the math on how long it takes for 64KB of window to get ackknowledgements. Either you have to modify TCP's window sizes, or use multiple TCP sessions, or use UDP with some kind of reliable-transfer application built over top of it instead of TCP.
Bill Stewart
New Fast-Compression-only CPR http://preview.tinyurl.com/dy575ks
Windows 2000 won it 3 years ago: Windows 2000 Sets New Internet Performance Record
What OS was it this time around?
(Seriously not flame bait, I'm curious.)
A speech...
For comparison, the TeraGrid backplane, running between hubs in Los Angeles and Chicago, is supposed to have a capacity of 40 Gb/s. No speed records yet; they're just sending the first test packets.
That's about 3000 kilometers. Assuming lightspeed transmission, there could theoretically be something like 40 or 50 megabytes of data at a time in transit.
So what's with this Internet 2?
People keep talking about this creatively named network and how fast it is.
Here's my question (please forgive or mod me down for this one, but I figure this is by far the best place to ask AND it's on topic)
What's the big deal? is this network using a different protocol, does it have BIG FAT network links causing it to be fast? why the big excitement over it? - I have no doubt you could acheive the same thing on the "internet 1" so to speak with a big enough link (s) ? right.
Is it a fibre only network, or it uses ip v6, or it has strong servings of beefcake? - what's the deal for the layman please.
Internet 2, why the fap?
...how's the ping time? Being on a semi-congested cable modem (with roommates who don't play online games, but do download), I seldom care about bandwidth... I'm usually griping about the 100 to 1000ms ping times that keep my awesome FPS skills in check.
:)
In other words, when do we get the quantuum packet transfer?