TY - GEN
T1 - HNFFS
T2 - 11th IEEE Non-Volatile Memory Systems and Applications Symposium, NVMSA 2022
AU - Pan, Yanqi
AU - Hu, Zhisheng
AU - Zhang, Nan
AU - Hu, Hao
AU - Xia, Wen
AU - Jiang, Zhongming
AU - Shi, Liang
AU - Li, Shiyi
N1 - Publisher Copyright:
© 2022 IEEE.
PY - 2022
Y1 - 2022
N2 - NOR flash, a type of non-volatile memory technologies, embraces its new age of IoT due to its execute-in-place (XIP) feature. Generally, there are two representative file systems designed for NOR flash: Journaling Flash File System 2 (JFFS2) and Spi Flash File System (SPIFFS). They suffer from either slow mount time, heavy foreground garbage collection (GC) overheads, poor read/write performance, or inefficient Wear- Leveling (WL), and thus they are unfriendly to IoT devices. To overcome the above limitations of existing NOR flash file systems, we propose and design a Harmony NOR Flash File System (HNFFS): (1) We introduce Erasable Sector Summary (ESS) to reduce mount time by leveraging NOR flash's fast read capabilities. ESS deploys summary nodes to avoid scanning the whole NOR flash when mounting the system. (2) We propose Adaptive Garbage Collection (AGC), which allows the NOR flash file system to adaptively trigger GC in advance and thus reduces the heavy Foreground GC overheads. (3) We introduce Mergeable Tree (MT) to decrease the DRAM usage for indexing while improving write performance by merging the sequential small-data write. (4) We provide Random Static Wear-Leveling (RSWL) to efficiently perform WL by combining the high performance Random WL (RWL) and the reliable Static WL (SWL). Experimental results suggest that HNFFS outperforms the state-of-the-art NOR flash file systems. Moreover, the evaluation of RSWL also shows the optimal tradeoff between performance and wear evenness.
AB - NOR flash, a type of non-volatile memory technologies, embraces its new age of IoT due to its execute-in-place (XIP) feature. Generally, there are two representative file systems designed for NOR flash: Journaling Flash File System 2 (JFFS2) and Spi Flash File System (SPIFFS). They suffer from either slow mount time, heavy foreground garbage collection (GC) overheads, poor read/write performance, or inefficient Wear- Leveling (WL), and thus they are unfriendly to IoT devices. To overcome the above limitations of existing NOR flash file systems, we propose and design a Harmony NOR Flash File System (HNFFS): (1) We introduce Erasable Sector Summary (ESS) to reduce mount time by leveraging NOR flash's fast read capabilities. ESS deploys summary nodes to avoid scanning the whole NOR flash when mounting the system. (2) We propose Adaptive Garbage Collection (AGC), which allows the NOR flash file system to adaptively trigger GC in advance and thus reduces the heavy Foreground GC overheads. (3) We introduce Mergeable Tree (MT) to decrease the DRAM usage for indexing while improving write performance by merging the sequential small-data write. (4) We provide Random Static Wear-Leveling (RSWL) to efficiently perform WL by combining the high performance Random WL (RWL) and the reliable Static WL (SWL). Experimental results suggest that HNFFS outperforms the state-of-the-art NOR flash file systems. Moreover, the evaluation of RSWL also shows the optimal tradeoff between performance and wear evenness.
KW - Embedded systems
KW - Flash file System
KW - NOR flash memory
UR - https://www.scopus.com/pages/publications/85140955504
U2 - 10.1109/NVMSA56066.2022.00012
DO - 10.1109/NVMSA56066.2022.00012
M3 - 会议稿件
AN - SCOPUS:85140955504
T3 - Proceedings - 2022 IEEE 11th Non-Volatile Memory Systems and Applications Symposium, NVMSA 2022
SP - 14
EP - 19
BT - Proceedings - 2022 IEEE 11th Non-Volatile Memory Systems and Applications Symposium, NVMSA 2022
PB - Institute of Electrical and Electronics Engineers Inc.
Y2 - 23 August 2022 through 25 August 2022
ER -