| Summary: | 碩士 === 國立臺灣海洋大學 === 資訊工程學系 === 97 === In current networks, the peer-to-peer network applications have received much attention. The most popular application is content distribution techniques. In peer-to-peer network, the system performance is depending on the behaviors of each peer. The selfish peers use resources but are willing to contribute, therefore impact the other peer’s download and hurt the whole network system. The peer-to-peer network attempts to reduce the number of selfish peers by employing the Tit-for-Tat strategy. To counterstrike the selfish peer, the upload amount is controlled so that it never exceeds the amount of download. It has been proved that this strategy can reduced the selfish behavior, but the constraint may be too strict to reduce the willingness of content delivering between peers, decrease the system performance, even the entire network might be crashed.
This paper studies a tolerable Tit-for-Tat strategy such that the difference between upload and download is only required to be less than a tolerance. We focus on the analysis of the download time and Seed workload, based on the number of participating peers and the number of pieces. We then derive some theoretical bounds on these performances. Compare the derivation of the results and the other strategies, proving that our strategy can efficiently download the file, reduce the Seed workload, and improve the whole system efficiency.
|
|---|
