Bittorrent updating nodes

Our session length data can be approximated with an exponential function and a better fit is obtained with the sum of two exponential functions. We also looked at the changes in number of users in different countries between 20.Note that since we only have data for 9 months in 2012 at the time of writing, these are not the final numbers, but they can be considered to show the trend.

The effects of real-world events can be clearly reflected in Bit Torrent System.We can see that Korean and US nodes are more likely to have longer session length, whereas Chinese users have typically shorter than average sessions.Generally, about 50% of the nodes have session length less then 200 minutes, while 3% have session length longer than 9000 minutes. We used the trust-region algorithm in Matlab to perform a Nonlinear Least Squares Regression fit to obtain these curves.The wrong image that the public had on Bit Torrent is partly due to the fact that academia failed to provide a sound measurement on this arguably the largest active overlay in the world.In the project, we first identify a systematic error in previous works measuring the number of nodes in DHT-based Bit Torrent networks (e.g., Mainline DHT, KAD, Vuze) and present the cause behind it, which is for measuring the size of Mainline DHT.