The new peering system is based on a Bittorrent Enhancement Protocol titled Canonical Peer Priority. This places increased importance on reducing the number of network hops between peers, which means that users will tend to prioritise peers who are geographically closer to them on the internet. The system also processes connections in a different manner as new - and potentially faster peers - enter a swarm. While not a complete reimagining of peer connections within a bittorrent swarm, Canonical Peer Priority acts as a more intelligent and focused selection system for making peering choices. The system also potentially has benefits for filesharers in reducing the impact of certain DDoS techniques as those attempting to poison a swarm with fake files are quickly removed from peer lists.
Scrutiny analysts undertook a range of download tests to offer evidence of the difference made by the changes to peering. More than fifty torrent files were downloaded from a popular tracker, comprising a range of popular content with a large number of seeds to those showing much poorer health and with fewer or no seeds. All testing was made on a standard 8MB ADSL home connection. At the one, two and three minute mark the current download speed was recorded. Comparisons were made between an older version (3.3.1) of uTorrent and the newest 3.4 release which introduced Canonical Peer Priority.
Testing found a noticeable difference in the average download speed between the old and new software. At the three minute mark, uTorrent 3.4 was around 38% faster in download speeds on average, showing an average throughput of 536.3kB/s versus 386.6kB/s for uTorrent 3.3.1. The highest percentage improvement was in the poorly-seeded releases where speeds improved by 86% while the newest and popular torrents (mainly newly released television episodes and new feature films) increased by 26%.
However, it seems that the higher transmission rate at the three minute mark is accompanied by slower initial transfer speeds when connecting to well seeded torrent files. At the one minute mark, downloads through the older client outperformed those using the new Canonical Peer Priority. If the new method of peering conducts a more complicated analysis of a swarm before attempting connections, it makes sense that this behaviour might be observed: it is likely to take longer to initiate connections than the previous randomised ‘first come, first served’ method of peering, especially when dealing with a well seeded file with lots of potential incoming connections. By the two-minute mark, uTorrent 3.4 had again established a significant download lead when accessing the most popular torrents.
Overall, the changes mean that a 1.4GB file (a typical size for a newly released feature film) completed a little over two hours faster using the new client than the old client on the 8MB/s connection. On a faster internet connection, download speeds may be slightly enhanced. The improvements noted are believed to have less effect where peers are plentiful and fast anyway - for instance, when using a private tracker. The slower startup may have an effect, particularly with impatient users, when attempting to access smaller files such as individual music tracks or ebooks. However, the trade-off of quick downloads and faster overall speeds is likely to be met with enthusiasm by the wider torrenting public.