A GEANT Leap Forward In Networking For Research

An anonymous reader contributes: "A research backbone network interconnecting more than 30 countries, through which hundreds of universities can exchange traffic, with a backbone running at 10 Gbps, born on the 1st of December. Yes, it exists, and this research network is not even in the U.S.! GEANT is a european initiative which has just come online, so if you're a student in Europe, you may have noticed a significant change in your downloads speeds since last week. You can even check its weathermap! Well, obviously backbone links are still unused ... but that shouldn't last long, once people notice the sheer amount of bandwidth."

Re:And as you can see, it's not using even...

[gowen]

What does that mean? It's not even using up, in almost all cases, any more than a 1Gbps line would be using. Take a look at all that blue on the map. It seems to signify that this was a waste of time and money.

Well the old European backbone was creaking slightly, so you can either upgrade incrementally to keep slightly ahead of demand, or oversupply now in the knowledge that in the next 5-10 years demand is going to keep going up and up.

Sounds like they made the right choice to me.

Re:And as you can see, it's not using even...

[Anonymous Coward]

"What does that mean? It's not even using up, in almost all cases, any more than a 1Gbps line would be using. Take a look at all that blue on the map. It seems to signify that this was a waste of time and money"

No, what would be a waste of time and money would be if it was at 100% traffic - the whole point about building a network like this is that it will cope with researchers' increasing demands for bandwidth for years to come. Of course traffic's low to start with, because people have been living with much lower bandwidth for years and don't suddenly start sending loads more data the second a new backbone appears. The bandwidth will be used when it's required, not when it's available.

Re:Yay!

[Psiren]

Not for academic institutions. Before I left my previous job, I was connected to JANET, the Joint Academic Network. I could quite regularly download at over 1 megaBYTE/s from other universites. Granted many sites in the US were still slow, but my local Debian mirror was shit hot ;-)

link

[Marcus Brody]

There was interesting article about this a few weeks ago in the gaurdian newspaper [guardian.co.uk].

Although it's pretty thin on technical details, it does provide some insight into some of the questions people are posting, such as why they need all this bandwidth, why the US arent part of the project etc.

btw, OC192 is (essentially) 10 Gbps

[tcyun]

I prob should have mentioned that OC192 is essentially 10Gbps in my earlier post. This means that Internet2 will be equiv (in terms of backbone speed) to GEANT in the near term. (You can read the PR [internet2.edu] about the Internet2 upgrade if you are interested.

Re:not all that much?

[kawlyn]

Dude that 10Gbps, the G meaning Giga.

Re:Doesn't really sound like that much...

[cnkeller]

This naturally makes me wonder what sort of backbones exist on the North America network, because I never have a problem downloading at 220KB/second, so I presume it must be pretty extraordinary.

There is a program called pathchar [caida.org] which seems to do a pretty good job of characterizing pipe size. I've used this to monitor my DSL bandwidth; PacBell has a 45Mbit line heading out of it's DSLAM's (at least in my area). It was designed to be used with symmetric connections, my DSL line (1.5/128) reports like 330K, but otherwise it's a good start at measuring paths.

From my office to microsoft's ftp servers I was easily able to determine that the slowest link is our T1 bewteen the ISP's T3 and our 10Mbps interface on our external router.

Some Perspective

[bjtuna]

This addresses fundamental routing issues, so my apologies to most of you, however I think some of this crowd needs some clarification (albeit a simplified version):

To all those who are posting such things as "now all I need is fiber to my home" or "I wonder if the Slashdot effect can saturate it" or "how come my ping times to it are so slow?":

You should know that hosts on these networks are generally a mix of globally- and non-globally-accessable. Meaning, many POPs that are "hooked up" to some high-speed initiative like vBNS or Abilene also have "commodity links." Commodity links are normal T3s, etc that are hooked up to a commercial ISP. This makes the site multi-homed, and helps minimize the amount of non-research-related traffic being sent over the high-speed links, because if you want to look at www.cnn.com from, say, a vBNS-connected box, it'll go over the commodity link instead of vBNS.

So the answer is, yes: the Slashdot effect can probably affect GEANT's web site because the Slashdot effect would flood their commodity link. On the other hand, if you were at a GEANT node... good luck trying, and enjoy the pings :)

-Brian
[e-mail] [mailto]

Re:It /does/ have a purpose

[Anonymous Coward]

Yes, No and Kinda - LHC doesn't come online until 2005/6 so at the moment the network is just being stress tested for something larger. However there are loads of projects at the moment in things like Broadcast quality video over IP (the Access Grid) which typically has three or more parallel connections out to multiple sites, the genetics and met. people are wanting to use it for data transfer (and at 8GB a file with a couple of hundred thousand files you see the problem).

Yes its quiet at the moment but the unis break for xmas in the next week so the researchers and overseeing exams etc for the undergrads, if its still that quiet in mid jan I'll be amazed and when the real stuff starts happening in march it'll really be flooded.

Some thoughts...

[jd]

First, why only 10Gbps? Lucent have 3 Tb long-distance optic fibre, and the primary cost of cable is in putting it into the ground, NOT the hardware.

Second, someone complained that they're only using a tiny percent of the bandwidth. Uhhh, the idea is to have SPARE capacity on a network. The three-way hook-up between Russia, Britain and the USA, for tele-surgery becomes actually practical for more than just extreme "he's very rich, but hasn't a hope in hell" cases. We might start seeing multi-national virtual operating theatres, capable of making use of a far wider range of skills than ever before possible.

IMHO, spending a few Euro more on slightly higher-quality fibre, and a few more frequencies of laser, is peanuts in terms of the total cost of a project like this, but offers the potential for fantastic endeavors that might actually benefit people.

The existing Internet would be fine, for most things, if it weren't loaded down with prawn and spam. However, it is, and we have to accept that. We also need to accept that the SERIOUS work on the Internet eats bandwidth for breakfast. When you're into real-time remote operation of a nuclear particle accelerator, online surgery, high-speed train emergency braking systems, etc, you really can't afford dropped packets, let alone serious lag.

Sure, AOLers can handle lag, just fine. What difference does an extra few minutes make, in a 2-hour download of a pirated DVD? Why the hell should they care about packet collisions or TCP retransmits?

But there are plenty of people, for whom a single packet collision could also be the last, if it happens at just the wrong moment. When you start talking about conditions like this, you absolutely need massive bandwidth. In fact, you really need three times that*.

(*It's a rule-of-thumb that network lag becomes significant, once you exceed one-third of the network's capacity. The odds of some form of data corruption, at that point, become too high to do even basic scientific work. You REALLY want the network to stay around the 1-5% region, for the high-end stuff.)