How Efficient Peer-to-Peer Video Streaming Could Be?

We consider the problem of broadcasting video streams to a large number of users through P2P overlay, which is emerging as a killer application after P2P file sharing. Considerable research has been conducted in this area in recent years. But most of them are experimental or on a particular algorithm/implementation. The lacking of a general model prevents formal studies on P2P streaming. In this paper, we present some interesting and important results of our work on the theoretical aspect of P2P streaming. The main contribution of this paper includes (1) proposing a general probability model making the analytical analysis and evaluation of P2P streaming possible; (2) with the help of the proposed model, proving that the upper bound of the efficiency for any P2P streaming networks is 1-(1/2k), where k is the average number of neighbors each peer knows about; (3) obtaining the condition to reach that upper bound and showing how to use it to guide real design; (4) showing that P2P streaming is quite different from P2P file sharing. It is not always safe to apply the experience obtained in one context to the other. In particular, we demonstrate that unlike P2P file sharing, P2P streaming does not always benefit from a wide piece diversity (i.e., the total number of the available pieces across the entire network at a given time).