A novel hot data identification mechanism for NAND flash memory

Hot data identification plays a very important role in NAND flash memory because it can improve the efficiency of garbage collection and decrease the degree of wear leveling. Existing hot data identification mechanisms have drawbacks in terms of their memory-space overhead and the true identification of hot data. To address these problems, this paper proposes a novel hot data identification mechanism. This mechanism mainly consists of kernel density estimation and a hot degree function. The kernel density estimation is used to build the kernel density function of the read and write operations by monitoring them in the NAND flash memory. That is, the kernel density function can be used for preliminary estimation of the probability distribution of the hot and cold data in NAND flash memory. After preliminary estimation of the kernel density function, the hot degree function is introduced to accurately identify the hot data in the NAND flash memory. Experimental results show that the proposed mechanism performs better than existing hot data identification mechanisms in terms of the false identification ratio between hot data and memory-space overheads.