Admittedly, “X percent of the Internet is P2P” makes for a great headline and is an effective way to generate publicity. It would be dishonest to deny this was the reason ipoque started its Internet studies in 2006. It is, at the same time, a great opportunity to demonstrate the own capabilities. If the equipment can detect and measure it, it is able to control it. Other equipment vendors have released similar studies, and analyst companies even try to make money with such studies. In the past two months only, three Internet studies with a particular focus on P2P traffic have been published, and ipoque’s Internet Study 2008 will be released by the end of the year.

Personally, I appreciate this diversity, even if the results are sometimes contradictory. These apparent discrepancies are sometimes used to argue that everything is just plain marketing and ipoque and all the others are lying to sell their products.

There are a number of reasons why the numbers are so different:

1. The measurement sites:
   How many business vs. residential users are represented by the measurement? Also, application usage patterns vary greatly with region and country.
2. The measurement methods:
   Deep packet inspection in combination with behavioral analysis in my opinion is the most accurate method to determine what kind of traffic is passing a measurement point. The quality of the detection method is crucial for the overall results. The error margin of the measurement lies in the range of the unknown traffic and provides a good estimation of the quality and credibility of study. Therefore, we always publish the amount of this unclassified traffic. And, PLEASE DO NOT trust someone who claims 100% detection rates. This is even more naive than believing in a 100% secure nuclear power plant.
3. Representativeness:
   No Internet study can be statistically representative. To conduct a statistically perfect study, a large number of measurement points had to be selected randomly. This is impossible. For these Internet studies, we need to rely on available data sources. ipoque is asking all its customers to contribute statistical data generated by the traffic management equipment installed in their networks. This is a difficult process involving a lot of persuasion and trust. No ISP wants to find itself in the same position as Comcast earlier this year.

Of course, I am convinced our numbers and methods are correct, otherwise we would not publish our studies. Just one note of caution, though, especially to journalists covering this topic: resist the urge to pick out only those numbers that appear to prove a certain point you want to make. At the same time, readers of such articles should always consult the original source for the complete picture!

So stay tuned, the 2008 Internet Study will have some interesting results.

Tags: Internet traffic, P2P, study

This entry was posted on Thursday, November 6th, 2008 at 16:49:20 by Hendrik Schulze. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.