Design and performance of a wired viral network

Abstract

We constructed a network that fully distributes access to audiovisual information. The information is apportioned among a family of machines. We call it Viral because it can scale in an ad hoc way and the addition of new nodes adds to the overall network capability. The network uses multiple multicast for distribution, acquires content from the broadcast television system, and makes viewing video content a more user-centric activity. In this thesis we address the performance of this network in comparison with other ways of providing the same spatio-temporal diversity of access to a body of work. We first provide some theoretical estimate of the capacity, and then we show how the network we built approaches those limits. This is done in terms of a presumed distribution of what the user wants.