Non-volatile Memory: Flash and Cache
summary
A set of talks about non-volatile memory technologies focusing on practical system-level problems: LLSM presents a lifetime-aware wear-leveling scheme for LSM-tree deployments on NAND flash; a paper on slotted paging proposes a low-overhead slotted-paging metadata scheme for NVM-based file systems; MAID-Q uses a Q-learning model to reduce tail latency in systems combining embedded flash with SMR disks; a WIP describes short-term lifetime prediction methods for TLC 3D NAND used as hot-data storage; another WIP explores prediction-driven fine-grained LDPC read strategies to improve high-density flash read performance; and ExpCache (WIP) introduces an online-learning cache replacement policy tailored for non-volatile memory.
AI-written summary · about