We develop Gibbs sampling based techniques for learning the optimal contentplacement in a cellular network. A collection of base stations are scattered onthe space, each having a cell (possibly overlapping with other cells). Mobileusers request for downloads from a finite set of contents according to somepopularity distribution. Each base station can store only a strict subset ofthe contents at a time; if a requested content is not available at any servingbase station, it has to be downloaded from the backhaul. Thus, there arises theproblem of optimal content placement which can minimize the download rate fromthe backhaul, or equivalently maximize the cache hit rate. Using similar ideasas Gibbs sampling, we propose simple sequential content update rules thatdecide whether to store a content at a base station based on the knowledge ofcontents in neighbouring base stations. The update rule is shown to beasymptotically converging to the optimal content placement for all nodes. Next,we extend the algorithm to address the situation where content popularities andcell topology are initially unknown, but are estimated as new requests arriveto the base stations. Finally, improvement in cache hit rate is demonstratednumerically.
展开▼
