Streaming Capacity in Peer-to-Peer Networks with Topology Constraints

Peer-to-peer (P2P) systems have shown enormous potential in delivering live streaming content for both IPTV and video conferencing. The fundamental capacity limits of the supported streaming rate by P2P, however, remains unclear, especially with topology constraints. We develop the first unifying framework and approximation algorithms that can compute the streaming capacity of a P2P network over arbitrary overlay topology, number of streaming sessions, and peer selection constraints. Converting a combinatorial multicast tree-selection problem into a linear programming (LP) problem, we show how a primaldual method can be used to handle the difficulty of exponential number of LP variables arising from the combinatorial nature of the P2P streaming capacity problem. Our proposed algorithms for constructively computing P2P streaming capacity establish a benchmark of any P2P system performance. Their effectiveness is demonstrated in simulations based on statistics of uplink capacities of actual Internet hosts.