Online File Caching on Multiple Caches in Latency-Sensitive Systems

Guopeng Li, Chi Zhang, Hongqiu Ni, Haisheng Tan

Abstract: Motivated by the presence of multiple caches and the non-negligible fetching latency in practical scenarios, we study the online file caching problem on multiple caches in latency-sensitive systems, e.g., edge computing. Our goal is to minimize the total latency for all file requests, where a file request can be served by a hit locally, fetching from the cloud data center, a delayed hit, relaying to other caches, or bypassing to the cloud. We propose a file-weight-based algorithm, named OnMuLa, to support delayed hits, relaying and bypassing. We conduct extensive simulations on Google’ trace and a benchmark YCSB. The results show that our algorithms significantly outperform the existing methods consistently in various experimental settings. Compared with the state-of-the-art scheme supporting multiple caches and bypassing, OnMuLa  can reduce the latency by \(14.77\%\) in Google’s trace and \(49.69\%\) in YCSB.