Enabling Adaptive Video Streaming for BitTorrent-Like Video Systems

Abstract

BitTorrent (BT) is one of the most efficient file sharing mechanisms in Internet. BitTorrent can distribute large files efficiently by using the shared bandwidth of the downloading peers. Using the rarest file first piece selection algorithm to download the data, BitTorrent’s downloaded data pieces are stored in an out-of-order way. Therefore, users can not watch a movie before the whole video file is downloaded. In order to support the video streaming for BitTorrent-based systems, we propose several mechanisms to improve the original BitTorrent protocol: (1) frame selection algorithms named decoding order frame first (DOFF) and rarest I frame first (RIFF); (2) an adaptive streaming retrieval scheme to allocate the available bandwidth adaptively for video streams. The adaptive streaming retrieval scheme periodically checks the streaming buffer to have knowledge of downloaded frames, and then allocates the available bandwidth for adapting frame downloading jobs with DOFF and RIFF frame selection algorithms. As a result, users can have smoother video playout experience than original BitTorrent protocol.