A Flash-Based Cache Optimization Strategy

Flash-based key-value caching is becoming increasingly popular in data centers. By caching most of data on SSDs, caching systems can eliminate a large number of time-consuming requests to back-end data storage and provide low-latency services. According to research, data redundancy is prevalent at all levels of storage systems with duplicate data taking up cache space and reducing the logical space for caching. Moreover, optimizing the performance of flash-based key-value caches is challenging due to the unique technical limitations of flash devices. In this paper, we propose a cache optimization strategy called PRCache, which designs a two-stage garbage collection scheme combining data deduplication techniques and replacement algorithms. We implement and evaluate PRCache, and experimental results show that PRCache improves the hit rate by 14.8% and reduces the average access latency by 13.7%.