Congestion Avoidance Social-Aware Caching Strategy in Cache-Enabled Heterogeneous Networks

Abstract

In cache-enabled heterogeneous networks (CHetNet), social-aware caching strategies are generally designed under the assumption that the user's sharing willingness (SW) is matched with its sociality, which indicates the probability that users in the real world build social connections with others. In practice, however, social-aware caching will make severe congestion if high-sociality users have weak SW. It is hard to handle the challenge since it is difficult to jointly quantify the impact of sociality, SW, and caching strategy on congestion. In this paper, we study the congestion avoidance caching strategy to maximize throughput capacity, considering user's sociality and SW. Specifically, the congestion mechanism in CHetNet is captured by efficient betweenness (EB), which is defined as the number of content delivery paths through a node. Aided by EB, it is interestingly revealed that popular contents should be preferentially cached in users with high SW and low sociality. Inspired by this, we propose a caching strategy to optimize the offloading ratio between layers and further tune caching probabilities to avoid local congestion. Simulation results show that the proposed strategy outperforms the compared sociality-preference schemes. Moreover, it is shown that the large caching buffer size may conversely decrease throughput capacity due to congestion.