File Aware Wear Leveling for PCM-based Mobile Consumer Electronics

Phase Change Memory (PCM) is considered as one of the most popular candidates to replace flash memory in mobile consumer systems. PCM has many superior performance characteristics, including non-volatility, byte-addressability, low access latency and power consumption. However, it also suffers from finite program counts like flash memory. Prior researches used PCM as a black box, and implemented the wear leveling schemes in device controller, which failed to utilize file attributes in host side and result in poor efficiency of wear evenness. In this paper, we propose a file aware wear leveling algorithm (called FAWL) for PCM-based storage system in mobile consumer electronics. FAWL is designed in the host side, which combines file attributes and statistical information of PCM. It exploits rich attributes of files to divide files into different categories and distribute them in suitable pages to avoid extra swap overhead. In addition, by utilizing an adjust management in FAWL, the wear imbalance can be greatly mitigated. Experimental results show that FAWL effectively improves the lifetime of PCM compared with existing wear leveling algorithms, including random swapping, start-gap and segment swapping.