An evaluation of piece-picking algorithms for layered content in Bittorrent-based peer-to-peer systems

In this paper the performance of layered piece-picking algorithms for Bittorrent-based peer-to-peer systems is evaluated and compared to traditional single layer solutions. In a Bittorrent-based peer-to-peer system, the piece-picking algorithm needs to ensure that the pieces of an audiovisual content are received in time to ensure smooth playback of the content. For layered content provided in multiple qualities, the task of the algorithm becomes more complex, as it has to consider the layer of the pieces to be fetched in addition to the deadline. Thus, the goal of the layered piece-picking algorithm is to ensure that the best possible quality under the given network conditions is received while the playback is not disturbed. To illustrate the advantages of such a layered piece-picking algorithm, its performance is evaluated and compared to a single layer piece-picking algorithm. The results show that layered piece-picking algorithms can significantly improve the quality in terms of peak signal-to-noise ratio.