Adaptive Switch on Wear Leveling for Enhancing I/O Latency and Lifetime of High-Density SSDs.

Jiaojiao Wu,Jun Li,Zhibing Sha,Zhigang Cai,Jianwei Liao
DOI: https://doi.org/10.1109/tcad.2022.3197340
IF: 2.9
2022-01-01
IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems
Abstract:NAND flash-based high-density solid-state drives (SSDs) commonly have limited write endurance, as a single block becomes unavailable after a finite number of program/erase cycles. The (static) wear leveling algorithms migrate cold data to more worn blocks for yielding an even distribution of wears in SSDs. However, migrating cold data must delay normal I/O processing, and unnecessary wear leveling operations may damage SSD endurance as each operation causes a program/erase cycle. This article proposes an adaptive switch on wear leveling (called ASWL), to boost I/O performance and lifetime for high-density SSDs. The basic idea of ASWL is to adaptively switch the threshold function for intelligently triggering wear leveling operations at different stages of SSD lifetime. To this end, we build a mathematical model determining the wear leveling threshold condition with a dynamic manner, by considering the real-time factors of I/O intensity and the wear difference of SSD blocks. Through a series of emulation experiments on several realistic disk traces, we show that the proposed ASWL mechanism can not only greatly improve I/O performance but also noticeably extend the lifetime of high-density SSDs, in contrast to existing wear leveling methods.
What problem does this paper attempt to address?