The Ono project, a product of Northwestern University’s AquaLab group, promises to accelerate your BitTorrent downloads to unheard-of levels: in average environments, Ono claims gains of 31% in download speed, while in higher-bandwidth environments users may see download speed improvements of 207% on average – and up to 833% if calculating by median!
With Comcast’s recent tussle over BitTorrent and its supposed crippling effect, it’s perfectly reasonable to believe that Ono would devastate ISPs’ infrastructure – but in fact the effect is opposite: Ono, which exists as a plugin for the supreme Azureus BitTorrent client, steers your computer into preferring network nodes closest to you, so that a given torrent is more likely to seek out nearby computers for downloading and uploading – providing the torrent’s swarm is large enough.
While geolocation certainly isn’t new – it’s seen use, much to my annoyance, in a lot of internet advertising as of late – the manner in which it is used is not conducive to BitTorrent sessions. Typical geolocation exists by comparing IP addresses to known real-world signatures, resulting in a rough approximation of an IP address’ physical location. Note that it returns a physical location, not a network one: my neighbor’s computer may be 100 feet away from my computer, but if she’s on another ISP her machine might as well be on the side of the country.
Ono works instead by seeking out the closest network machines, that is, computers that are the fastest accessible over a user’s ISP’s network, relative to them. It checks this by taking cues from content distribution networks: companies like Akamai already have thousands of servers placed around the world, and they exist solely to offload bandwidth-intensive downloads from their client’s web sites. When a user requests a file hosted by Akamai, he or she is redirected transparently to the fastest available server, and that response is often one designed to send users to a server that is most convenient for his or her ISP – resulting in a vastly accelerated download. Ono captures that redirection information, then compares the result to the results of its peers. Azureus is instructed to connect to nodes Ono feels to be closest, resulting in higher-quality, lower-latency BitTorrent downloads.
Yesterday, I decided to give Ono a try. After installing the plugin, I unleashed it upon the massive, 24-bit 96khz version of Trent Reznor’s latest, coincidentally released for free on the same day. The download weighed in at 1.06 gigabytes, and was previously chugging along, on average, at a weak 60 to 80 kilobytes per second. Shortly after installing Ono – it didn’t even require me to restart my copy of Azureus – it began probing servers belonging to Akamai and Limelight, collecting statistical averages of how many times each of their servers were seen. After checking back a short while later, I noticed my download speed consistently stayed at 150 kb/sec, and it remained at this speed until the torrent completed and switched to seeding mode.
Anecdotally, it appears that Ono works as advertised. My modest 3-megabit DSL connection isn’t nearly large enough to see the high-end of Ono’s claims, but over my “average” internet connection, Ono did an excellent job of turning a trickle, if you will, into a torrent.
The interesting thing is that the Ono project predates the Comcast/BitTorrent controversy by at least a few months, which means that the project’s maturation seems to have nothing with the joint agreement between Comcast and BitTorrent, Inc. This, in turn, means that we can hopefully look for even more improvements in the future: perhaps Comcast’s/BitTorrent’s optimizations will synergize with AquaLab’s, and the will be stratospheric – without ISPs wailing and moaning every step of the way. Here’s to hoping.