Live seeding: Performance bounds of seeders for P2P live streaming

Seeders (peers that do not request anything hut contribute to the system) allow to leverage the capacities of a P2P system. While seeding is a natural idea for fllesharing or Video-on-Demand applications, it seems somehow counter-intuitive in the context of live streaming. This paper aims at describing the feasibility and performance of P2P live seeding. After a formal definition of "live seeding" and efficiency, we consider the theoretical performance of systems where the overhead is neglected. We then propose a realistic overhead model and extend the results for this model. The performance of a single seeder and a set of seeders are considered, as it is not always possible to perfectly aggregate individual efficiencies. In details, we provide an explicit upper bound of seeders' achievable efficiency in a P2P system with linear overhead. We also propose and study two simple mechanisms that allow to deploy a live seeding architecture while handling seeders aggregation, providing near-optimal seeding.