Research on write optimization of NVRAM memory management system based on decision tree and LSM-Tree.

Non-volatile random memory has the characteristics of fast read and random storage of SRAM and DRAM, and the non-volatile characteristics of magnetic disks, which has become one of the current research hotspots of memory. However, the writing speed of the non-volatile random access memory is relatively slow, and how to improve the writing speed of the non-volatile random access memory has become an important issue for research. A large number of scholars have designed a special memory file management system by studying the structure of non-volatile memory, but its write speed is still lower than that of DRAM. This paper proposes a classification write operation algorithm, which uses decision tree algorithm to intelligently classify write requests, combines LSM-Tree ideas for memory management, and uses DRAM as a cache to write to NVRAM through asynchronous write back operations, greatly improving NVRAM write speed.