A Bandwidth-Aware Scheduling Strategy for P2P-TV Systems

P2P-TV systems distribute live streaming contents by organizing the information flow in small chunks that are exchanged among peers. Different strategies can be implemented at the peers to select the chunk to distribute and the destination neighboring peer. Recent work showed that a good strategy consists in selecting the latest received chunk and a random neighboring peer (latest useful chunk, random peer). In this paper, leveraging on the idea that it is convenient to favor those peers that can contribute the most to the chunk distribution, we propose to select the destination peer with a probability proportional to the peer upload bandwidth. We show that the proposed scheme has a limited sensitivity to cheating peers that maliciously declare higher bandwidth than they actually have. Considering the overlay topology, we evaluate both systems in which nodes have fixed degree and systems whose overlay setup takes into account the actual peer bandwidth by assigning more neighbors to peer with higher bandwidth. We evaluate the performance in terms of delay percentiles and loss probability and evaluate the achieved improvements. Simulation results considering scenarios with up to 10,000 peers shows that the proposed schemes significantly outperform the traditional ones, so that the chunk distribution delay drops to less than 2 s from about 12 s.