CIF:Small:Towards practical coded caching

Iowa State University, Ames IA

Investigators

Abstract

Content caching plays an important role in facilitating large scale content distribution over the Internet. Traditional caching techniques typically store relatively popular content in the local memory (or cache) available at the users, or at the edge of the network. If a given user request can be serviced from the cache (instead of the server), the overall network traffic is reduced. This research will investigate practical issues in the deployment of coded caching, which is a technique that promises huge reductions in caching network traffic, albeit under potentially restrictive assumptions. This will pave the way for the adoption of coded caching in large scale video and audio streaming websites thus improving the efficiency of the national network infrastructure. Minimizing per-user delays is one of the main aims of caching. For example, a given video-on-demand user wants his/her video playback to start as soon as possible. Moreover, it is not reasonable to expect that several users want to watch videos at the same time. The investigator will study practical algorithms that a server can use for responding to asynchronous user requests. The algorithms will continue to leverage the rate reductions of coded caching while meeting user-specified deadlines. Another issue with coded caching is that it assumes that the files at the server can be subdivided into a large number of sub-files. Specifically, this number grows exponentially with the number of users and is impractically large even for systems with fifty users. The investigator will study the of design coded caching schemes where the subpacketization level is manageable, yet significant rate reductions are possible as compared to conventional schemes.