Slashdot Mirror
Mapping the Internet Evolution
Shire writes "Science magazine is running a story on the DIMES project, which has ventured to map the structure and evolution of the Internet (PDF) using open source distributed clients in the style of SETI@Home and such. DIMES has already collected more than 40 Millions measurements which resulted in some nice pictures and several scientific presentations. Those who use traceroute may find it a useful (and colorful) alternative."
http://mirrordot.org/
:-)
All the links mirrored.
"Your effort to remain what you are is what limits you."
Mirrordot Science magazine is running a story on the DIMES project, which has ventured to map the structure and evolution of the Internet (PDF) using open source distributed clients in the style of SETI@Home and such. DIMES has already collected more than 40 Millions measurements which resulted in some nice pictures and several scientific presentations
If you pay your taxes you support terrorism!
More informational links about DIMES that aren't slashdotted...well, at least not yet:
f
d f
- http://dawn.cs.umbc.edu/INFOCOM2005/shavitt-sl.pd
- http://dawn.cs.umbc.edu/INFOCOM2005/shavitt-abs.p
Beer is proof that God loves us and wants us to be happy. -- Benjamin Franklin
Mirrordot mirrors of the pictures, karma-whore free: one two
Opera 8 suppports SVG by default, I think
I have my own little internet mapping project ( http://tr.meta.net.nz/ ) which is designed in a similar way (people run traceroute nodes on their machines and information is merged together to provide pretty graphs). I wrote it because people would say "I can't get to this site, can anyone else get to it?" This lets you type in the hostname of a machine and it will take lots of traceroutes from around the Internet and merge them into a graph that you can use to figure out which particular segments of the Internet can/can't access it and where they all get tripped up. Or you can see that you're going via an international route, where as almost everyone else is going across a local exchange point. Also as it has AS# information on it you can determine who's fault it is.
r osoft.com.png . google.com.png o ot-servers.net.png
tr produces a "small" section of the Internet (it doesn't map the entire thing) but it produces it in a way that can be interpreted by anyone savvy in network administration. It's mostly based in NZ (as thats what I care about, and thats where I have contacts where people are happy to run tr nodes) but it does show how the NZ Internet works extremely well, and provides reasonable detail to the rest of the Internet.
Some interesting examples:
Microsoft: http://tr.meta.net.nz/output/2005-04-08_20:08_mic
Google: http://tr.meta.net.nz/output/2005-05-16_10:14_www
F root server anycast: http://tr.meta.net.nz/output/2005-05-16_10:16_f.r
Well, we're working on a linux version now. The agent is not totally java, since java does not support pinging and tracerouting and other stuff like that, so a small part of it is native. A linux version will be coming soon so check the site back in a couple of weeks.
In a nutshell, it's kinda showing you connectivity and relative size of an ISP. What it's showing with the red spheres is that there are a few Autonomous Systems (AS) that have lots (hundreds or thousands) of inbound and outbound links to other ASes. Think UUNet or Sprint, alot of people are connected to them, as well as some ISPs. Then you have those smaller (tier-2, the orange to blue skittles) ISPs that are connected to UUnet and also connected to tens of other ISPs. All the way at the edge, you have people who are single homed only to their ISP, but they are not a transit network, meaning they don't send traffic from one AS to another. Any traffic on their link is either from their network or to their network. And there are lots and lots of those, they are the little purple skittles on the outside.
The lines represent the connections between these ASs. All the real connections arent there, the entire background would most likely be completely black if they were.
So it's showing that there are some really huge ISPs that form the "backbone" of the internet, and lots of smaller ISPs and thousands of even smaller ISPs. When you see a line that does not connect to/from the center, that's a connection between a tier-2 ISP to another tier-2 ISP. Usually this is done as a backup mechanism, sometimes it can also be done to more effectively get traffic to another network. For example, if UUNet charges you per byte, and they can get to anyone, you would think that you could just send all your traffic to UUnet. But you might find that a large percentage of your traffic is going to networks that are directly attached (peered) to Level3's network, and Level3 charges less per byte. It might make sense then to peer with Level3 in addition to UUNet, or completely with Level3. I'm leaving some stuff out here but this is the general idea, and what the picture is showing you.
Incidentally, there is also a concept in the internet backbone called "hot potato routing". If I am in LA with Sprint and sending traffic to NY to a UUnet customer, Sprint will do everything it can to put the traffic on UUNet's network while in LA. In otherwords, if this traffic is for a UUnet customer, it won't be sent across Sprint's network from LA to NY. Sprint drops it on UUNet's network at the first opportunity. The routing protocol for the backbone, BGP, has more than 10 different metrics it looks at to decide which route to take, and several rules to follow to decide what is the "best" path. Best is not always shortest or fastest once all the metrics are taken into account. So if you thought the internet was a big harmonious cooperative effort, guess again!