Breadcrumb
Waiting on Distributed Content
Fri 15 March 2013, 15:40
Jean-Francois Chamberland
Texas A&M University
Organisers: Nick Whiteley, Feng Yu
ABSTRACT
Information mixing is an established and valuable paradigm for distributed storage. For instance, error-correcting codes can be employed to introduce redundancy in data located over a network. A properly designed coding strategy increases the reliability of the distributed storage system. Yet, in large environments, the amount of traffic necessary to maintain or rebuild a distributed storage system can be sizable. Recently, network coding has emerged as a means to alleviate the adverse effects of repair traffic and complex disk updates. A side benefit of these advanced techniques is a reduction in the time necessary to restore the integrity of a corrupted system. Information mixing can also be used to improve the performance of content delivery networks with caches. Coding enables users to fetch information from arbitrary caches, which in turn facilitates load balancing and reduces latency. This talk discusses the potential benefits of network coding in the context of content delivery networks. A queueing perspective is adopted, with a special emphasis on completion times. The potential benefits associated with specific mixing strategies are quantified, and design guidelines for content distribution systems are identified.