Dynamic Memory Management for Hybrid DRAM-NVM Main Memory Systems

Non-Volatile Memory (NVM) is a promising alternative to DRAM in embedded systems due to its attractive characteristics. However, the endurance of NVM is limited and influences greatly its adoption in embedded systems. To reduce the write times on NVM for hybrid DRAM-NVM main memory systems, this paper proposes a new memory management algorithm, called DMA, which measures write pressure of programs and allocates the appropriate memory (NVM or DRAM) for these programs. According to the experimental results, the proposed algorithm can reduce the number of writes on NVM by 40%~60% compared with the case without this algorithm. Moreover, the experiment results also indicate that DMA can decrease NVM write times by 50% on average, regardless how many programs are running at the same time. These results prove that DMA is able to extend NVM lifetime for hybrid DRAM-NVM main memory systems.