RandMap: Wear Level for Phase Change Memory Based on Layer-based Random Mapping

Phase change memory (PCM) is a new kind of nonvolatile storage. Compared to traditional memory, it has many advantages, such like lower-energy use, higher density and better scalability. PCM is supposed to replace DRAM and become the next generation of memory in embedded systems. However, PCM still has some disadvantages, such as the limitation of writing times. Although some approaches have been proposed to relieve these disadvantages, they can not well resist the malicious wear attacks. In this paper, we propose a practical approach named RandMap to achieve wear level in PCM. RandMap separates PCM into layers, and each layer uses its own random mapping table to achieve wear level when converting the logic address to physical address. When the writing time of a layer reaches the threshold, its random mapping table will be updated immediately. RandMap can implement wear level and enhance the ability of resisting malicious wear attacks, and it is available for large-size PCM. Our evaluation on SPEC2000 and SPEC2006 benchmarks shows that RandMap can improve the wear-level effect to 77.2% compared to previous approaches, and it can support the PCM size up to 16GB.