摘要
使用主动探测BitTorrent tracker服务器的方法,对4个BitTorrent tracker服务器的资源共享流行度和资源下载流行度进行了测量与建模,并使用χ2拟合检验法对模型进行了检验。发现资源流行度具有严重的不均衡性:最流行的20%资源具有约60%的共享,而最流行的10%资源具有60%以上的下载。进一步解决了目前P2P资源流行度研究中存在的Zipf模型是否适用的争论,得出以下结论:①Zipf模型和Zipf-Mandelbrot模型都适合于对资源共享流行度进行建模,而Zipf-Mandelbrot模型比Zipf模型建模的效果更好;②Zipf模型不适合于对资源下载流行度进行建模,而Zipf-Mandelbrot模型具有完美的建模效果。
A detailed study on the BitTorrent sharing objects popularity was presented. The objects-sharing popularity and objects-downloading popularity of four BitTorrent trackers was measured and modeled by means of actively probing the tracker server, and the model was tested by Z2 fitting test. The experiment results indicate that the peer-to-peer sharing objects popularity is very unbalanced, that is, there are almost 60% of total share copies for the most 20% popular objects and there are above 60% of total downloads for the most 10% popular objects. Moreover, the experiment results also indicate that the objects-sharing popularity can be well modeled by Zipf distribution and perfectly modeled by Zipf-Mandelbrot distribution, and that the objects-downloading popularity can not be modeled by Zipf distribution and can be perfectly modeled by Zipf-Mandelbrot distribution.
出处
《通信学报》
EI
CSCD
北大核心
2009年第3期66-74,共9页
Journal on Communications
基金
国家重点基础研究发展计划("973"计划)基金资助项目(2007CB307100)~~