Similarity-aware popularity-based caching in wireless edge computing

Mobile edge computing (MEC) can greatly reduce the latency experienced by mobile devices and their energy consumption through bringing data processing, computing, and caching services closer to the source of data generation. However, existing edge caching mechanisms usually focus on predicting the popularity of contents or data chunks based on their request history. This will lead to a slow start problem for the newly arrived contents and fail to fulfill MEC's context-aware requirements. Moreover, the dynamic nature of contents as well as mobile devices has not been fully studied. Both of them hinder the further promotion and application of MEC caching. In this backdrop, this paper aims to tackle the caching problem in wireless edge caching scenarios, and a new dynamic caching architecture is proposed. The mobility of users and the dynamics nature of contents are considered comprehensively in our caching architecture rather than adopting a static assumption as that in many current efforts. Based on this framework, a Similarity-Aware Popularity-based Caching (SAPoC) algorithm is proposed which considers a content's freshness, short-term popularity, and the similarity between contents when making caching decisions. Extensive simulation experiments have been conducted to evaluate SAPoC's performance, and the results have shown that SAPoC outperforms several typical proposals in both cache hit ratio and energy consumption.