CIF: Small: Fountain Coding Theory for Content Sharing and Multimedia Broadcasting in Overlay Networks

Colorado State University, Fort Collins CO

Investigators

Abstract

Abstract Multicast applications, such as content sharing and multimedia broadcasting, is the single most important class of applications over the internet. Multicast applications assume overlay network models, which consist of relevant nodes connected by abstracted virtual logical links. While scalable multicasting in such a network requires error-correction coding, classical coding approaches are highly inefficient due to the fact that virtual links do not enjoy known statistical channel models. To overcome this challenge, this project will develop a systematic coding theory based on a new fountain communication model for efficient multicasting in overlay networks. In fountain communication, the source node encodes a message into an infinite number of packets, a destination decodes the message after the number of received packets exceeds certain threshold. The investigator will first develop a fountain coding framework for end-to-end multicast transmission over virtual network links. For delay nonsensitive traffic, fountain codes will be developed to achieve ideal rate-error performance with a linear coding complexity. For delay sensitive traffic, fountain coding will be integrated with flow control algorithms to ensure timely information delivery over channels with arbitrary distortions and erasures. The objectives are characterizing fundamental performance limitations and developing practical coding schemes. The project will then investigate multiuser cooperative communication for multicast applications in node-capacitated overlay networks, where the sum upload rate of each node is kept below a predetermined bound. For delay sensitive communication, the objective is to overcome the challenge that network nodes may randomly access or disconnect from the network. For delay nonsensitive communication, the objective is to overcome the "selfish-peer-node" challenge where a peer node disconnects from the network whenever its desired messages become fully decodable.