Robust Network Transmission and Storage Using Coding
Abstract
This project focused on transmission and storage of information in networks, and considered robustness to adversarial errors, packet losses, link failure, mobility and topology dynamics. It established fundamental limits on performance (capacity, reliability and delay), as well as practical coding schemes and optimization techniques. Among our results, we developed network error correction theory for networks with non-uniform link capacities, non-multicast and a priori unknown number of errors. We applied this theory to design coding schemes for robust key distribution and streaming. We designed codes for error estimation, universal multicast codes robust to changes in network size and number of receivers, and error detection codes for distributed storage. We proposed efficient methods for obtaining network capacity bounds, characterized the impact of the failure of a single link on capacity of some families of networks, and showed equivalence between the Shannon capacity for saturated sources and the stable capacity of networks with probabilistic message arrivals. We characterized optimal resource allocation for maximizing the probability of data recovery under probabilistic access or failure of storage nodes, and optimizing transmission delay in disruption tolerant networks.
Document Details
- Document Type
- Technical Report
- Publication Date
- Aug 09, 2013
- Accession Number
- ADA590765
Entities
People
- Tracey Ho
Organizations
- California Institute of Technology