Wired magazine recently asked us for information about P2P Filesharing traffic in North America.  In response, we provided a figure from our Fall 2010 Global Internet Phenomena report which shows that P2P Filesharing was 19.2% of aggregate traffic during peak period on North America’s fixed access networks.

Wired also approached another vendor with experience in traffic identification, and it seems that there is a sizeable discrepancy between the two responses that the magazine received and reported on (8% was reported by Arbor Networks). What could cause such differing figures?

The difference is that Sandvine’s traffic identification technology breaks BitTorrent into four sub-protocols, which are summed together as “BitTorrent” in our Global Internet Phenomena reports.  When BitTorrent first debuted, we differentiated between “BitTorrent” (hereafter referred to as “BitTorrent (regular)” and “BitTorrent (UDP)”.  As encryption techniques appeared, we responded by recognizing this traffic as “BitTorrent (encrypted)”.  Finally, when uTorrent popularized the uTP protocol, we introduced a “BitTorrent (uTP)” recognizer.  On North America’s fixed access networks, BitTorrent breaks down as shown in the pie chart – figure 1.  This breakdown is fairly consistent throughout the day, although the absolute levels of BitTorrent vary as shown in figure 2 below.

BitTorrent typically represents about 90% of all P2P Filesharing traffic; consequently, in the Fall of 2010 BitTorrent accounted for roughly 17.3% (90% of 19.2%) of aggregate traffic in North America’s evening hours.  Breaking this down into the sub-protocols yields the following shares of total traffic:

• BitTorrent (regular) – 6.3%
• BitTorrent (uTP) – 6.7%
• BitTorrent (encrypted) – 2.9%
• BitTorrent UDP – 1.4%

Of these constituent parts, “BitTorrent (regular)” and “BitTorrent (UDP)” are trivial to identify, whereas the encrypted and uTP varieties require very sophisticated traffic identification techniques.  Perhaps this reality can explain the discrepancy between the numbers provided by Sandvine and those provided by other organizations.  It’s possible that our P2P Filesharing numbers (which included all varieties of BitTorrent) were being compared against only the easily-detected protocols.  In fact, if you sum together our numbers for “BitTorrent (regular)” and “BitTorrent UDP”, you get a figure of 7.7%, quite close to the 8% reported by Arbor Networks.

Figure 1

Figure 2