BitTorrent在多檔案下載環境中的效能分析

Abstract

BitTorrent(BT)是一種點對點(peer-to-peer)的檔案傳輸協定，由於其獨特的設計在傳輸效能上有著非常優異的表現，因此常常被用於在網路上散佈大型檔案或用於高畫質的影音串流傳輸，BT也是目前世界上流量最大的網路協定之一，也因此吸引了不少學者針對BT這個協定進行了大量的研究。 其中BT最為驚人的一項特性也就是其下載時間不受使用者進入系統的速率的特性，造就了其獨特的可擴展性(scalability)，更讓BT一時之間成為了被大家爭先研究的對象。但這些研究多數是爭對單一檔案下載(single-file download)情況下的研究，儘管有少部分關於多檔案傳輸環境的研究，但仍然缺乏對於多檔案下載(multiple-file download)環境統一的整理和分析。並且根據觀察實際世界中的BT使用情形，超過85%的使用者其實同時間都在傳輸複數的檔案也就是都處於多檔案傳輸的情況下。 我們的研究發現在網路頻寬的效用(utilization)很高的情況下，所使用不同的多檔案傳輸方式，例如MFMT(Multiple-File-Multiple-Torrent)、MFST (Multiple-File-Single-Torrent)或是針對MFST的情況下增加所使用的上傳連結(unchoking slot)數量等等多種方法再無論檔案大小是否均一的情況下效能都是相同的。另外我們也在實驗中發現都整體網路的頻寬較小時，使用者進入系統的速率越高的時候，系統平均的下載時間會變得更長。

BitTorrent is a peer-to-peer file sharing protocol; due to its unique mechanism, it has excellent file transport performance, so it is widely used to distribute large amounts of data across the Internet and to stream high-quality video. BitTorrent is also one of the largest traffic consumers on the Internet, attracting various research on this protocol. One of the most amazing characteristics of BitTorrent is that the average time a peer stays in the system is not related to the entry rate of peers into the system. This indicates that BitTorrent has great scalability. This characteristic attracts even more research on this protocol, but most of it focuses on the theme of single-file downloads. Research on multiple-files downloads exists, but there is still a lack of rigorous, organized research on the topic. According to real-world traces, over 85 percent of BitTorrent users download multiple files concurrently. Our research found that when the bandwidth utilization is high, different multiple-file-download approaches─such as MFMT(Multiple-File-Multiple-Torrents)，MFST (Multiple-Files-Single-Torrents) or MFST with extended unchoking slots have similar average download times, regardless of whether of file sizes are unique or not. We also found that the entry rates of peers will actually affect the average download time of peers when the system is operating in a low bandwidth environment.