Distributed proactive caching in a group of selfish nodes

Abstract

The paper considers the problem of proactive caching (or replication) of content in a collection of caches which can form a group and collectively satisfy content requests of their users (clientele); such caches can be the ones associated with close-by base stations. Under significant user similarity, avoiding caching the same content in multiple locations within the group and leaving room for more content to be cached within the group, can yield significant performance improvement compared to the case of isolated, greedy local, caching behavior by the nodes. The work considers a generic content-provisioning cost structure that jointly incorporates objective (i.e., distance-based) and subjective (social/behavioral-based) cost elements and studies the resulting non-cooperative game, proposes a simple content placement implementation, presents results and discusses the impact of the subjective cost element.