Coded Caching: Global vs Local Content Popularity

In the coded caching problem a file server with N files is connected through an error-free broadcast link to K users, each equipped with a local cache. The placement and delivery phases can be jointly optimized to create simultaneous multicasting opportunities among users that request different files. We examine the model where we have multiple user classes and users belonging to different user classes have different demand popularities. We develop placement schemes based on either the global popularity of the files or the local popularity of the files and for each of these placement schemes, we examine delivery schemes based on either the user perspective or the file perspective. We analyze the server transmission rates for these schemes and identify parameter regimes where schemes working with local content popularity profiles do better than those utilizing only the global popularity.