On System Time Analysis for BitTorrent with Sharing Ratio Enforcement
Yuh-Jzer Joung, Evan Chang · 2013
BitTorrent's built-in mechanism can effectively encourage peers to upload while they are downloading, but it lacks a mechanism to incentivize peers to continually upload to benefit others after they have completed downloading. As such, many private BitTorrent sites have enforced a {sharing ratio} on their members to demand the minimum amount a peer must upload with respect to the amount it has downloaded. In this paper we study how sharing ratio enforcement affects download time by addressing the following problem: {Given a flash crowd of peers downloading a file, what is the time required for the last peer to finish downloading?} We propose two models to estimate the download time. The first model is simpler in the sense that it does not need to know peers' actual download rates during the downloading process. However, it needs to assume that every peer can fully utilize its uplink capacity. This assumption may not hold when there are much more peers to upload than to download, a scenario that is not uncommon in a system with sharing ratio enforcement. The second model lifts the assumption by extending an existing model for estimating peers' actual download rates in a system with no sharing ratio enforcement. However, estimating peers' actual download rates becomes very complex when the number of peers of different bandwidths increases.