McCormick Magazine

Clearing the path

Identifying nearby computers to speed P2P traffic

Email::

bustamanteIn the early days of the Internet, traffic was low, and users could spend a lazy day driving through scant message boards and minimalist web sites. But with the explosion of music, video, and other file sharing, the Internet is becoming clogged with the virtual semitrailer traffic of downloading and uploading from across the globe.

Peer-to-peer (P2P) file-sharing services, which connect individual users for simultaneous uploads and downloads directly rather than through a central server, are reported to account for as much as 70 percent of Internet traffic worldwide. That level of use has led to a growing tension between Internet service providers (ISPs) and the P2P file-sharing services their customers use. And it has driven ISPs to forcefully reduce P2P traffic at the expense of unhappy subscribers and the risk of government investigations.

Now researchers at McCormick have discovered a way to ease that tension: Ono, a unique software solution that allows users to efficiently identify nearby P2P clients. The software, which is freely available and has been downloaded by more than 250,000 users, benefits ISPs by reducing costly cross-network traffic without sacrificing performance for the user. In fact, when ISPs configure their networks properly, Ono significantly improves transfer speeds — by as much as 207 percent on average.

Ono was developed by Fabián E. Bustamante, associate professor of electrical engineering and computer science, and PhD student David Choffnes, and it has been deployed for the Azureus BitTorrent P2P file-sharing client. “Finding nearby computers for transferring data may seem like a simple thing to do,” says Choffnes, “but the problem is that the Internet doesn’t have a Google Map. Every computer may have an address, but it doesn’t tell you whether the machine is close to you.”

Worse yet, the simplest solution to finding computers that are close to you requires measuring the distance to most of them — an operation that is too costly and time consuming to be practical.

Instead, Ono — Hawaiian for “delicious” — relies on a clever trick based on observations of Internet companies like Akamai (Hawaiian for “clever,” incidentally). Akamai is a content-distribution network (CDN) that off-loads data traffic from web sites onto their proprietary network of more than 10,000 servers worldwide. CDNs such as Akamai and Limelight power some of the world’s most popular web sites and enable higher performance for web clients by sending them to one of those servers in close proximity. Using the key assumption that two computers sent to the same CDN server are likely close to each other, Ono allows P2P users to quickly identify nearby users.

Ono is different from other software applications that address the conflict between ISPs and P2P traffic (see, for example, the recently announced partnership between Verizon and P4P) because it requires no cooperation or trust between ISPs and P2P users. Ono is also open source and does not demand the deployment of additional infrastructure. Bustamante’s Aqualab research group made Ono publicly available in March 2007 and recently published code that makes it easy to incorporate Ono services into other applications. “The more users we have, the better the system works, so we’re just trying make it easy to spread,” says Bustamante.

Since news of Ono first hit the media (including being featured on the popular technology blog Slashdot.org) in May 2008, the number of users has grown from 130,000 to over 250,000. Most recently, the team has created and made available NEWS, a new plug-in for Vuze/Azureus BitTorrent clients that allows peers to cooperatively detect network problems and unfriendly ISPs. In just a couple of months and without any publicity, the new service has already been downloaded by more than 7,000 users.

To learn more about Ono (and other related source code), visit the Aqualab web site at http://aqualab.cs.northwestern.edu.

—Kyle Delaney