CCFlash: A Correlation-Aware Compression Approach in Flash Memory

How to improve the read performance and endurance of the flash memory has always been a critical concern. In this paper, we reconsider this problem from two independent angles: access correlation and data compression. We first demonstrate via real-world trace analysis that correlated chunks are often prevalent and of great significance to practical accesses in storage systems. We then present CCFlash, a correlation-aware compression approach. CCFlash first designs a lightweight yet effective algorithm to capture correlated chunks. It then sets forth to compress the correlated chunks for opportunistically reducing the write traffic and co-locating them in the flash memory. CCFlash finally carefully puts forward new read and write strategies for effectively incorporating the proposed correlation-aware compression designs. By reducing the write traffic via the data compression and leveraging the inferred access correlation in the read operations, CCFlash can shorten 19.5-47.3% of the read latency and eliminate 17.2-40.8% of the garbage collection operations, while in the meantime shortening 4.1-18.3% of the write latency.