Internet speed record quadrupled to 101 Gbit/s

[matcom]

<snip>

Los Angeles (CA) - An international team of physicists, scientists and engineers achieved a new speed record for long distance data transfer: During the Supercomputing Bandwidth Challenge the sustained data transfer was 101 Gbit per second (Gbp/s) between Pittsburgh and Los Angeles. This is more than four times faster than last year's record of 23 Gbp/s.

The group's "High-Speed TeraByte Transfers for Physics" record data transfer speed is equivalent to downloading three full DVD movies per second, or transmitting all of the content of the Library of Congress in 15 minutes, and it corresponds to approximately percent of the rate that all forms of digital content were produced on Earth during the test, the team led by the California Institute of Technology said.

The extraordinary achieved bandwidth was made possible in part through the use of the Fast TCP protocol developed by Professor Steven Low and his Caltech Netlab team. It was achieved through the use of seven 10 Gbp/s links to Cisco 7600 and 6500 series switch-routers provided by Cisco Systems at the Caltech Center for Advanced Computing booth, and three 10 Gbp/s links to the SLAC/Fermilab booth near Chicago.

The team hopes this new demonstration will encourage scientists and engineers in many sectors of society to develop and deploy a new generation of Internet applications.

The new record was submitted after a demonstration took place in the framework of the Internet2. The project, launched on October 1, 1996 today combines more than 200 universities and is considered an updated version of the NSFNet, the successor of origin of the Internet, the Arpanet. Scientists consider the Internet2 as their dedicated communication tool between universities as well as a test bed and playground to prepare new applications and technologies for the "common" Internet.

http://www.tomshardware.com/hardnews/20041129_153412.html

[lapfog_1]

1. Silly wabbits

Never underestimate the bandwidth of a FedEx truck full of tapes.
(or 747 or pick your own).

It's not the bandwidth, it's the latency! ( just how much is it
worth it to have the data available NOW versus tomorrow morning?)

[DS1]

2. Rich media, so developers can further bloat internet apps

Developers like me, of course.

[lapfog_1]

5. Nope, not hardly

There is so much "dark fiber" trunking capacity now that it's just
plain silly. your apps (or any rich media app) depends on last mile
bandwidth plus, likely, some latency issues... and that just isn't
happening. Most folks here, even with "broadband", are still in the
sub 1 Mbit/sec range, and almost everyone is sub 10 Mbit/sec.

So these experiments are entertaining, and if some research
institutions can afford it, might be useful for shipping large
files around (10 GB or bigger). But really, why ship such things
around? It's far cheaper to replicate them and send them via
UPS (that's what I did at NASA, and the economics haven't changed
that much since I left). So, you are left with time critical
large files which are shared by large institutions or corporations
which need the type of service this experiment shows...

MRIs or something maybe... but those just aren't that large.

[DS1]

7. A ridiculous conclusion

[lapfog_1]

9. Uh, what is a ridiculous conclusion?

I don't know who you are, but I've been working on internet
"things" (like the first generation IMPs) for almost 30 years
now. While I was at NASA (for over 10 years) I was in charge
of developing 3rd generation internet systems and applications
(IPv6). Not to mention developing storage systems for one of the
largest supercomputer centers in the world.

I also ran the group that developed mftp (multi-stream ftp)...
etc, etc.

100 Gbit/sec bandwidth (4 DVDs per sec) is simply not useful in
any real sense. And certainly not economically justified for real
applications (at least those that are not classified, and even
classified seems to be a reach, but I'll admit that I'm not privy
to all of those, but they don't use the "internet" anyway).

I created one of the fastest IP apps in the country, driving a
Virtual Wind Tunnel cave (virtual reality) using actual solution
sets from a cluster of supercomputers... and even that was in the
single digit Gbit/sec transfer range.

So, care to point out WHAT is ridiculous in my last post?

[DS1]

12. No user will ever require more than 640K of Ram

Just because you can't imagine why people would need that much data throughput, doesn't mean it won't be required in the future. People don't have fat pipes for the last mile, currently. Fiber will be eventually in everyhouse, if not citywide gigabit wireless connections.

[lapfog_1]

13. That's not what I said.

And the big difference is that 100 Gbit/sec is 100,000 times faster
than the current "last mile" fat pipe. So what application developed
today needs that. As for fiber to the home or gigabit wireless...
Not in the next 5 years. And not because of any technological reasons.

It's because the folks who might role out that technology simply won't
make those types of investments in infrastructure anytime soon. They
are too busy trying to recoup from the bloat of over purchase from
the last 10 years of Internet hype. They basically got talked into
paying WAAAY too much for the internet backbones that they acquired,
counting on the exponential growth AND fairly constant pricing of
MB/sec (neither of which occurred of course... so they went broke,
and the ones that didn't are now gun shy).

Eventually, you are correct. 10 years... maybe longer.
And this little "dick measuring" exercise that they engaged in
is simply not that hard to do, nor, as it turns out, all that useful.
For the applications that they can think of right now, I submit
that the cheaper alternative (and you DO want you tax dollars used
correctly, right?) is FedEX or UPS. The bandwidth is just HUGE!
And the cost per Mbit/sec is really, really cheap. No routers, no
dedicated OC-48s, blah blah blah. Which is why I made this same
argument 5 years ago for mass data exchange within NASA (really
just between two NASA centers that dealt with massive data
exchanges). Low tech, but effective.

[DS1]

14. Right, and it is a dicksize contest and I'm not surprised

Cisco had something to do with it. Yet, even with last mile pipes being what they are now and what they'll be for the next ten years, having bigger trunks isn't a bad thing, especially if newer technologies help keep network saturation below 20%. Once resources are made available, people will figure out how to fill it. I wouldn't mind seeing full frame 30 fps realtime video tomorrow, coupled with just about every household having this capability trunk size is going to need to make these leaps and bounds.

Then again, the money could probably be better spent on an entirely new TCP scheme, or something of that nature.

[meegbear]

3. Whoo-Hoo! Now we can download porn faster!

God bless them science peoples.

[Moonbeam_Starlight]

4. Um, WHOA.

[BigMcLargehuge]

6. ARGHGHGHGHGH PET PEEVE ALERT!!!!

it's not speed, it's RATE.

the speed of information traveling through copper, irrespective of the amount of information pushed, is the speed of light minus the coefficient of copper. The speed of information traveling through glass fiber is the speed of light minus the coefficient of the fiber core.

1k of information travels just as fast as 10000000000000k over a copper, fiber, wireless link.

Speed doesn't change... RATE CHANGES!!!!

The RATE of information transfer was 101 GB/s.

AAAAAAAAHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHH!!!!!

Okay, I feel better now. Please go about your business, nothing to see here.

[lapfog_1]

10. And, again, it's not the bandwidth

(which you correctly point out is the RATE of data delivery)...
it's the latency that is really important. And they have done
nothing to fix that (other than IGNORE it for things like window
acks, etc).

[underpants]

8. GREAT! So I can send THIS to everyone even FASTER!!!

[BlueJazz]

11. I remember a DU'er posting once that he was having problems...

...with his Dial-up service and he was runing about 27 BYTES per second. He said it took about 9 minutes to load the DU home page.

I think I posted: "Geez pal...I think that breaks a World Record" :)

[Xithras]

15. Umm, thati's INTERNET2, not the public Internet

FYI, Internet2 is a private communications network between a bunch of the major universities and is completely disconnected from the Internet that we know. It is not possible for private parties to get access, and even if you could somehow sweet talk a connection, the membership fees are in excess of $20,000 per year.

In other words, this is a nice technical leap, but don't look for any big jump in your Internet performance in the near future.

Besides, this performance jump occurred because it used a completely different protocol to transfer the files than the current Internet uses. In order to roll this out onto the public Internet, every computer, switch, and router connected to it would have to be upgraded to handle it. Something tells me that we wont see that happen anytime soon.

[Endangered Specie]

16. So how fast can I download a porno movie now?