FACC: A Flexible and Asynchronous Updating Strategy for Cooperative Edge Caching

Recently, cooperative edge caching effectively reduces redundant transmissions and user plane latency by deploying potentially requested videos in edge nodes. Many existing works focused on designing accurate and diverse video selection strategies. However, they ignored that the edge nodes' heterogeneous requirements of replacement timings also impact the caching performance. In this paper, we propose FACC, a cooperative caching framework in which each edge node can flexibly select variable replacement timings and asynchronously update cached videos. We formulate the asynchronous cache replacement problem as an integer programming problem with a semi-Markov property. To solve this problem, we propose HiMAPPO, a multi-agent hierarchical deep reinforcement learning algorithm. It has been validated that FACC can effectively improve the edge hit rate and significantly reduce replacement costs compared with existing approaches on a real-world dataset.\