A decentralized architecture for adaptive media dissemination

P2P (peer-to-peer) content distribution networks have become extremely popular on the Internet. Due to their self-organization properties, they suffer from the lack of control to balance the load among peers for contents of different popularity. We define and analyze a fully decentralized architecture to support replication of popular media content in a peer-to-peer network. We show that by exploiting local knowledge only, this method achieves a near-optimal behavior with respect to response time in a scenario with limited storage capabilities at peers.