Caching Salon: From Classical to Learning-Based Approaches

As a key technology in content delivery, content caching can not only reduce the load on the origin server to save the traffic costs for content providers, but also can enhance users' Quality-of-Experience (QoE) by placing popular objects in close proximity. In reality, Content Delivery Network (CDN) providers always try to design their caching strategies carefully to improve the hit ratio of requests. With the development of learning-based methods and their great successes in many fields, researchers are also trying to explore learning-based caching approaches in recent years. In this paper, we first review the classical caching approaches, most of which are based on recency, frequency and content size. Then we summarize the recent learning-based methods in content caching field. Finally, we discuss the potential content caching strategies in the future, and point out some open issues. Through the studies in this article, we wish it can be used as a reference for interested parties.