Please use this identifier to cite or link to this item:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/64140
Title: | 強制分享率下BitTorrent系統的效能分析 Performance Analysis of BitTorrent Systems with Sharing Ratio Enforcement |
Authors: | Evan Chang 張子捷 |
Advisor: | 莊裕澤(Yuh-Jzer Joung) |
Keyword: | BitTorrent,強制分享率,效能分析, BitTorrent,Sharing Ratio Enforcement,Performance analysis, |
Publication Year : | 2012 |
Degree: | 碩士 |
Abstract: | BitTorrent(BT)是目前最為廣泛使用的點對點檔案傳輸協定,它曾經長期占據網際網路流量第一名,直到最近才被Netflix超越。其迥異於其他點對點檔案傳輸協定的獨特機制達到了非常好的效率並兼具了成長性,也讓它成為檔案分享社群的不二選擇。但是其機制無法鼓勵使用者在下載完成後做為種子(Seeder)繼續上傳,因此許多社群便導入了Sharing Ratio Enforcement(SRE)這個外部機制來解決這個問題。
在本篇論文中,我們在探討SRE對於BitTorrent系統效能的影響。我們建構了一個數學模型,可以在Flash Crowd的情況下,預測一個不限數目且異質網路環境中BitTorrent系統完成下載所需要的時間。再來,我們結合了一個前人的模型,讓我們可以更進一步的預測當系統內的總上傳頻寬高於總下載頻寬時,這個系統完成的時間。以往的研究大都假設系統效能不會卡在下載頻寬上,但是在SRE機制下,系統內可能充滿大量種子,便有可能導致這個情況。最後我們針對上述情形,提出下載者可以策略性的選擇從何處下載,並以我們的模型驗證改變選擇下載來源的策略會影響系統完成的時間。 BitTorrent is the most widely used Peer-to-Peer file sharing protocol on the internet. It had been the largest consumer of internet traffic for years, until surpassed by Netflix in recent years. The unique mechanisms and the built-in incentives that distinguish BitTorrent from other P2P file sharing protocols result in excellent utilization and scalability, making it the top choice for file sharing communities. One of the problems of the built-in incentives is that it lacks the incentive to encourage peers to stay seeding after they finished downloading. So a lot of the communities have incorporated a mechanism called Sharing Ratio Enforcement to compensate it. In this paper we study the impact of the SRE mechanism on the system performance of the BitTorrent systems. We derive a mathematical model that is capable of predicting the time that all peers finish downloading under a flash crowd scenario and heterogeneous network with an arbitrary number of classes of peers. Furthermore, we incorporate our model with a model from a previous work by Liao et al. to make it capable of modeling a system that the system throughput is capped on the downlink instead of the uplink. Finally we propose that the leechers can strategically select from which peers to download when there are too many seeders in the system that the uploading throughput is larger than the downloading throughput of the system. We study how this modification influences the system performance by using our model. |
URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/64140 |
Fulltext Rights: | 有償授權 |
Appears in Collections: | 資訊管理學系 |
Files in This Item:
File | Size | Format | |
---|---|---|---|
ntu-101-1.pdf Restricted Access | 757.54 kB | Adobe PDF |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.