Slashdot Mirror
Internet2 Plus P2P Equals...
Bill, I'm lost in cyberspace... writes "News.com has an article up about a Direct Connected P2P network set up at universities which are on Internet2. This is majorly cool! More direct information is available at i2hub.com for those lucky enough to be located at a University with Internet2 access."
Why is there a special p2p for I2? Here at RIT we have I2 and it just works automatically. If you try to connect to an I2 computre the router does everything transparently. So whenever we connect to another college student with a p2p app I2 is used automagically.
The GeekNights podcast is going strong. Listen!
During my last year of University the network was so bogged down with p2p traffic (or so they told me) that it was a nightmare trying to download any kind of large file. Suddenly it felt like I was on dialup again. The only way to get a decent speed file transfer was from someone at another Internet2 connected school. My understanding was that any connection between 2 schools that were on Internet2 would automatically use the faster other pipe.
This is purely a social networking system rather than a technical one. If I'm on an i2-enabled conenction, and you're on an i2-enabled connection, then any direct connection between me and you over any protocol is going to route over an i2 bandwidth link rather than going out over the open Internet link between our two sites...
Really, this is like when the Starr Report against then-president Clinton, and all sorts of ISPs who don't do content mirroring did a mirror for that document, since it was long and going to be frequenly downloaded that day. By keeping that traffic local on their own network, their outgoing Internet line was freed up for other traffic.
Knowing who is closer to your network-wise, which isn't aways the shortest physical difference or lowest number of network hops, but the one who has the most available bandwidth on the path that leads from you to them and back, when given a choice between mirrors is always very useful.
So, really, i2hub's goal is to just point out where useful content is on i2 rather than change any routing tables...
I went to the website and clicked the link for the Mac OS X software. The title on the download page reads:
Direct Connect for Mac OS X 1.1.0
Author:NeoModus
Program Type:Adware
Interesting...
A common misconception is that this hub is "safe" because it is on I2. This is not so. People have been sued from using i2hub. No, I will not say who, but yes, I will say it's happened.
:-/
Filesharing is just bad. Of course I do it anyway. But if you look at your campus' bandwidth usage, some ludicrous number like 95% will be p2p traffic
I've been using I2Hub for a few months and the downloads aren't that fast, at least as someone who has been spoiled by the internet connection here at RPI. Downloads from a user at another college with i2Hub are usually in the range of 30-80 kb/s. For comparison, this is about the same speed that I get from a p2p app that let's you download from multiple sources, such as eMule. If someone I know off campus, such as a friend from home, downloads a file from me via AIM, they get speeds ranging from 150-200 kb/s. However, for ease of use and individual files, as well as a better community, I2Hub is pretty good.
I'm at Brown and before they capped the i2 bandwidth I was downloading movies on the i2hub at over 8 MB a second. I honestly couldn't really believe it--full length movies in under five minutes.
Ummm... Just to let you know, the reason that Internet 2 is faster is that it's completely separate lines, running on a 10 gigabit backbone, with many schools having a gigabit or higher connection to it. Yes, it's used for research, but most lines are less than 10% utilized, so I don't think it's having any effect on the speeds of the network.
The reason this would be such a bad thing is, while TCP has "fairness" built in, so that one connection doesn't monopolize the network, UDP has no such mechanism - in fact, what I described in the parent is the opposite of fairness: each connection tries to hog as much bandwidth as possible. This app would, effectively, cause Inet2 to grind to a halt.
There's a reason why network admins hate UDP apps.
dinner: it's what's for beer
Here's a better way to figure out how to get an abserdly fast connection rate...
First... broadcast a request for the file (specified by a hash value, just like BitTorrent does it.) to the local subnet. If anybody hears this request and responds with an "I've got it!"... bingo. There's no need to go to the external network. This file is already locally available, so copy the file using a LAN-protocol that assumes insane speeds and little packet loss. If nobody answers, you've wasted less than a kilobyte of local bandwidth and you can move on...
Now, send an HTTPS request for a list of the people known to be making the file available to the main tracker-server. However, there's a little trickery at this stage. Instead of just returning a flat list, the tracker-server notices the "apparent IP address" where the request came from. For those directly connected to the Internet, this is their true IP address... however, for those going through a NAT situation, this is the outwardly-visible IP address that speaks for the true client. The list is sorted by whatever server's apparent IP address most bitwise matches the requester's address. This leads to some interesting situations...
- If there's a server with the same apparent IP address as the client IP address... this means that the content is being offered up within the local network again, just not on the same subnet. There's no reason to involve the NAT any further... we just need to introduce two computers that are behind the the same NAT/Firewall to each other. Again, a local-area transfer protocol gets pulled out.
- Now we start getting to the level where we have to assume things. For example, it's rather safe to assume that a subnet whose IP space is one, two or three bits away from my subnet is most likely to be "somewhat local" to me. Those are the servers that get tested first for connection speed... and you keep going down the list until you have relable connections to enough servers to end up maxing-out your incoming bandwidth. At that point, you'll have your file shortly.
An interesting point is that this will end up detecting most i2-links because there's a cluster of colleges in the 128.x.x.x IP space because those were the first 3-octet ranges handed out and the colleges who were the first in line to get those were also the ones who were first to sign up for internet2. So, just by giving preference to those servers who are within the same first-eight-bits of IP space as you, you're more likely to find a faster link...
Port numbers have nothing to do with transfer speed.
symetrix. We are building a religion, a limited edition.
11 megs a second is 100mbit ethernet, it's not fast, compared to Internet2.
No, we've had a number of different DC hubs at RIT over the past year or two. They keep getting shut down, so new ones have to open up. Do you know why they get shut down in the first place? Because of morons blabbing to slashdot about it. Seriously, don't go off on slashdot bragging that RIT is some long-standing mecca of piracy (even if we were the subject of some major bust by the FBI two years ago). You might as well walk through Brooklyn announcing you've got $500 in cash on you, or bragging to some IRS agent about how you just cheated on your taxes. Posting this anonymously because I seem to have misplaced my flamesuit...