Performance Limits of Coded Caching under Heterogeneous Settings

Caching plays an important role in reducing the backbone traffic when serving high-volume multimedia content. Recently, a new class of caching schemes, called coded caching, has received a significant amount of interest. Coded caching can exploit new multicast capabilities even when multiple users request different pieces of content, and thus can further reduce the backhaul requirement. While most of the prior studies of coded caching have focused on homogeneous settings, the performance limits of coded caching under heterogeneous system settings is much more involved. In this talk, we present two recent studies along this direction. In the first part of the talk, we study the setting where the files are of heterogeneous popularity, and we derive lower bound and achievable bound of the expected transmission rate of coded caching that are at most a constant factor away. Unlike other existing studies, the constant factor that we derived is independent of the popularity distribution. In the second part of the talk, we study the setting when the files are of heterogeneous sizes, and we derive lower bound and achievable bound of the worst-case transmission rate of coded caching that are at most a logarithmic factor away. These studies reveal new insights on how to use coded caching to effectively deal with heterogeneity. In both cases, the techniques for deriving the tighter lower bound may also be of independent interest.