PM-based Persistent Key Value Stores: a Survey
Xuran Ge, Yang Liu, Lizhou Wu, Yang Ou, Zhiguang Chen, Nong Xiao · 2022 5th International Conference on Data Science and Information Technology (DSIT) · 2022
The advent of persistent memory(PM) technologies provides an opportunity for building persistent key-value stores by making the best of its data persistence and access performance close to DRAM. In recent years, researchers have proposed many key-value store designs, and these improvements come from the database and operating system communities. This paper particularly introduces the optimization schemes and improvements of various index structures, such as hash tables, B+ trees and LSM-Trees on PM. It discusses their highlights and trade-offs in write efficiency, data persistence, crash consistency and concurrency control. Finally, this paper provides researchers with optimization directions to cultivate persistent key-value stores deeply.