An active protocol architecture for collaborative media distribution

Abstract

This thesis embarks on distributing the distribution for real-time media, by developing a decentralized programmable protocol architecture. The core of the architecture is an adaptive application-level protocol which allows collaborative multicasting of real-time streams. The protocol provides transparent semantics for loosely coupled multipoint interactions. It allows aggregation and interleaving of data fetched simultaneously from diverse machines and supports the location and coordination of named data among peer nodes without additional knowledge of network topology. The dynamic stream aggregation scheme employed by the protocol solves the problem of network asymmetry that plagues residential broadband networks. In addition, the stateless nature of the protocol allows for fast fail-over and adaptation to departure of source nodes from the network, mitigating the reliability problems of end-user machines. We present and evaluate the algorithms employed by our protocol architecture and propose an economic model that can be used in real-world applications of peer-to-peer media distribution. With the combination of an adaptive collaborative protocol core and a reasonable economic model, we deliver an architecture that enables flexible and scalable real-time media distribution in a completely decentralized, serverless fashion.