Peer-to-peer live streaming: Optimality results and open problems

Many peer-to-peer systems have been deployed on a large scale to provide users with "live streaming", that is Internet-delivered real-time multimedia content, much as the traditional television service. All such systems are organized in an unstructured manner, with epidemic-style information exchanges: users decide who to interact with in an adaptive manner, on the basis of past experience, and decide which data blocks to exchange with their "logical neighbours" by relying on simple, local rules. It is therefore important to understand the performance limitations of unstructured approaches: if such designs can ensure live data diffusion at rates and delays that are competitive with those of more sophisticated approaches, then these must be the winning designs. In this paper we describe recent results which establish rate and delay performance optimality results for specific dissemination rules. We also formulate open problems concerning scenarios in which the efficiency of unstructured live streaming systems is still poorly understood.