Slashdot Mirror
Intenet2 Backbone Upgrades
An anonymous reader "Looks like Abilene, the backbone for Internet2 will join Canada's CA*Net3 and Europe's GEANT as one of the fastest research networks on the planet. According to this press release, Internet2 will be deploying 11 of Juniper network's freshly announced T640 platform. These puppies can cram 32 OC-192 (or 128 OC-48) interfaces into a single chassis. All in half a rack, too!" I'm
sure those students are very happy with their ping times. Meanwhile in the
real world... ;)
Without napster, do we really need all that bandwidth anymore? ;-)
Slashdot needs a speel cheeker...
Kramer
"What's this script do? unzip ; touch ; finger ; mount ; gasp ; yes ; umount ; sleep Hint for the answer: not everyth
"...and this is why I think it is very important to study the effects, upon international policy-making by semi-marginalized non-governmental stakeholders, of three-day Quake matches. I thank the comittee for their time."
Carousel is a lie!
"Designed for deterministic performance with 640-Gbps font-panel throughput and 1,280-Gbps rear-panel throughput"
That's a lot of bandwidth killed if someone trips on the power cord.
"A terrorist is someone who has a bomb but doesn't have an air force." -William Blum
The internet? that thing still around?
-Homer
No, Microsoft is not on Internet2.
:)
traceroute research.microsoft.com
traceroute to research.microsoft.com (131.107.65.14), 30 hops max, 38 byte packets
1 gallgtwy (134.231.4.2) 592.570 ms 40.421 ms 9.430 ms
2 gallgw (192.26.10.1) 0.557 ms 0.540 ms 0.459 ms
3 d3-2-1-1.a00.mclnva02.us.ra.verio.net (168.143.233.85) 1.308 ms 1.188 ms
[Lines deleted]
Verio is our Internet uplink.
If I go to UMD, my network goes through I2 with 1ms ping times.
traceroute www.umd.edu
traceroute to websrv1.umd.edu (128.8.10.105), 30 hops max, 38 byte packets
1 gallgtwy (134.231.4.2) 2.251 ms 2.226 ms 2.689 ms
2 gallgw (192.26.10.1) 0.870 ms 0.613 ms 0.488 ms
3 clpk-t3-1-3-2.maxgigapop.net (206.196.177.133) 1.490 ms 1.484 ms 1.570 ms
4 wash-umcp.maxgigapop.net (206.196.177.50) 5.203 ms 380.967 ms 8.777 ms
5 Vlan14.css-core-r1.net.umd.edu (128.8.7.193) 1.767 ms 1.666 ms 1.577 ms
6 websrv1.umd.edu (128.8.10.105) 1.792 ms 1.631 ms 1.604 ms
Yes, we are very happy with how fast our Quake 3 Arena games are...
Looks like Abilene, the backbone for Internet2 will join Canada's CA*Net3 and Europe's GEANT as one of the fastest research networks on the planet
According to this page at Geante,
An important element of GÉANT is the development of connectivity with equivalent Research Networks in other world regions. Connectivity is being consolidated with the existing equivalents of GÉANT in North America (Abilene, CA*net) and in Asia-Pacific (SINET, KOREN, SingAREN) and developed further between Europe and the Asia-Pacific, North American, South American and Mediterranean regions
a bunch of extra regions get connected as well.
Er - possibly because realistic timekeeping wouldn't be possible otherwise?
Despite the throughput of these lines their latency will still be high when talking about transatlantic distances.
As far as I understood it a lot of this bandwidth will be used for real time work as well as transfer of large amounts of data - for real time you need to know latencies.
A fast line in Gbps is not necessarily a fast line in pings - 6 million modems will give you a 192Gbps connection - but the ping times will be stupid.
Matt Thompson - Actuality - Insert product here.
I could hit 1meg a second to windowsupdate.microsoft.com
That must be fantastic -- imagine having to only spend a couple of hours downloading security updates.
You would expect the slashdot *editors* would have discovered the distinction between latency and troughput by now. 128 (or whatever) OC-12 running in parallell does not give you a lower ping-time than a single one. (unless your high ping is caused by congestion)
What it does is allow you to transfer more data. Consider this analogy: Sending a hundred postcards at once doesn't make your message get there faster, but it *does* give you space for a longer message.
Ofcourse Internet2 is also built to have low latencies, however the humongous bandwith doesn't contribute directly to this, except as in making congestion less likely.
DWDM would allow a single ring to cram anywhere from 32 x to 256 x the OC-192 capacity, on a single fibre (and on expensive equipment, that goes without saying :)
All major telcos/routers companies have nice DWDM offerings already today, and much more in their labs. Links: Nortel, Lucent, Cisco ...
But these P2P apps adapt (simply because they are evil) and we are already seeing increases traffic. So guess what? We have to buy more bandwidth. I wonder if Joe Taxpayer likes the idea that his pennies on the dollar toward education go for through bandwidth at a blackhole so kids can playu Quake instead of studying. We roll the 622Mbps link on July 1 with one of those badass Juniper routers ($80000) to boot.
Comparing it to Windows will be a moot point, since El Dorado is going to have a 40% larger code base than XP.
During the good old days of networking when I was at Virginia Tech, they had a pretty interesting setup. As far as I understood, VT used to sit on a NAP on the I2. The closer you were to the NAP, the fatter your pipe. There were some plans to open the NAP up for local residental access since most of the Blacksburg residents were students and faculty. I don't know if that was ever accomplished or not.
:-)
:-/
Anyways, before digressing, VT's outgoing pipe had two logical interface. Any packets bound to universites or other educational institutions that had access to the I2 via their local NAP points, would go through the then established oc-3. (The pipe might be fatter now). Any other packets that were bound for networks outside of these destinations were forwarded through the dual t-3 that was used for 'all other traffic'.
I onced did a traceroute to www.ucla.edu from a computer lab on campus during the middle of the day during the middle of the week and got amazing results. I found that there was only 8 hops between that desktop and the webserver that was in CA somewhere and all ping responses were less than 10ms. Talk about insane.
I believe other schools share the same network setup as VT and i wouldn't be surprised many of those once old pipes have now been upgraded to fatter ones. Then again, MCI does have a lot of dark fiber laid around the AMTRAK rails that has yet lit up.
However, despite with all this nice connection, I was recently told by several Virginia Tech on-campus residents that their connection has been capped up. I did some digging around and I believe that CNS is now capping the wall connections with the use of the catalyst 6500 catalysts from Cisco which I belive can limit network usage from reading all their marketing material... lol
Bottom line: Even if your organization or institutions had fat pipes to external networks, if your network capacity is limited from the point where you plug in your RJ45... don't expect to see blazing speeds).
BTW, as far as I know, they got the ports to the residents dorms set up to 10mpbs half duplex... ewwwww......
The router flows for some of the routers on Internet2 still show a lot of file sharing apps even on Internet2. Heres a break down for the LOSA router (I believe that's Los Angeles).
port flows octets packets duration
FastTrack 22.010 26.377 17.495 19.339
Gnutella 8.358 5.069 7.138 11.082
http 4.201 4.566 2.565 1.151
ftp-data 0.738 3.284 1.866 0.915
eDonkey-2000 0.896 1.132 0.769 1.111
ssh 0.428 1.063 0.753 0.337
Neomodus-Direct 0.591 0.706 0.823 1.057
51872 0.017 0.513 0.302 0.086
ftp 0.636 0.444 0.337 0.296
aol 0.139 0.428 0.302 0.291
bbh
I've been living on-campus at a canadian university for 2 years now and only recently discovered how amazing the research network is. (Our regular commercial line is really slow, slow enough to prompt the luckier, and geekyer, residents with TV to get cable internet in addition to the residence internet.)
If you check out the traffic graphs, you can see that well over half the traffic is kazaa. (click on application-bits)
http://205.189.33.73/www/flowscan/nrc.html
Taxpayers' dollars hard at work indeed! The cool thing is that at most times these nodes aren't anywhere near their maximum data transfers at any time that I check them. That's probably just because nobody really knows about it and only use it if they happen to connect to someone else on the network and their university has the routing setup correctly... Also, not all the universities in Canada I've connected to make full use of the network, some limit bandwidth to their users even on this "free" (gov't subsidized) network. From what I hear though, the free part will soon change and the universities/gov't offices will have to pay for it in the upcoming years, but right now it's basically free bandwidth for those on the network.
Actually, your traffic never even makes it to I2 at all. The "Maxgigapop" or "Mid-Atlantic Crossroads" is a regional aggregation point for I2 members. Both your school (which is in DC) and UMD (in Maryland obviously) connect to the MAX, and the MAX has a link to I2. But in your case, the traffic never has to go all the way to I2. Which explains the crazy-low ping times. The packets basically never even leave town (which is why regional aggregation points are good.) Try a traceroute to some schools on the west coast, and you will see ping times which have some measurable delay in them (due largely to the speed of light).