An Adaptive Buffering Scheme for P2P Live and Time-Shifted Streaming

Abstract

Recently, P2P streaming techniques have been a promising solution to a large-scale live streaming system because of their high scalability and low installation cost. In P2P live streaming systems, however, it is difficult to manage peers' buffers effectively, because they can buffer only a limited amount of data around a live broadcasting time in the main memory and suffer from long playback lag due to the nature of P2P structures. In addition, the number of peers decreases rapidly as the playback position moves further from this time by performing time-shifted viewing. These situations widen the distribution of peers' playback positions, thereby decreasing the degree of data duplication among peers. Moreover, it is hard to use each peer's buffer as the caching area because the buffer area where the chunks that have already been played back are stored can be overwritten at any time by new chunks that will arrive soon. In this paper, we therefore propose a novel buffering scheme to significantly increase data duplication in buffering periods among peers in P2P live and time-shifted streaming systems. In our proposed scheme, the buffer ratio of each peer is adaptively adjusted according to its relative playback position in a group by increasing the ratio of the caching area in its buffer as its playback position moves earlier in time and increasing the ratio of the prefetching area as its playback position moves later. Through extensive simulations, we demonstrate that our proposed adaptive buffering scheme outperforms the conventional buffering technique considerably in terms of startup delay, average jitter ratio, and the ratio of necessary chunks in a buffermap.