Hi-ZNS: High Space Efficiency and Zero-Copy LSM-tree-based Stores on ZNS SSDs

被引:2
作者
Liu, Renping [1 ]
Chen, Junhua [1 ]
Chen, Peng [1 ]
Long, Linbo [1 ]
Xiong, Anping [1 ]
Liu, Duo [2 ]
机构
[1] Chongqing Univ Posts & Telecommun, Chongqing, Peoples R China
[2] Chongqing Univ, Chongqing, Peoples R China
来源
53RD INTERNATIONAL CONFERENCE ON PARALLEL PROCESSING, ICPP 2024 | 2024年
关键词
ZNS SSD; LSM-tree; Space Efficiency; Data Migration;
D O I
10.1145/3673038.3673096
中图分类号
TP3 [计算技术、计算机技术];
学科分类号
0812 ;
摘要
The Zoned Namespace (ZNS) SSD is a newly introduced storage device and provides several new ZNS commands to upper-level applications. Zone-reset command is one of the ZNS commands to erase all the flash blocks within a zone. Since data is grouped and erased in zone units, ZNS SSDs are widely used in LSM-tree-based stores. However, the basic invalidated unit in LSM-tree is an SST/WAL file, which mismatches the erasing unit of a ZNS SSD. Placing different SST/WAL files in the same zone, LSM-tree on ZNS SSDs faces dramatic space amplification and extensive data migration problems. To solve these problems, Hi-ZNS aligns zone sizes with varying-sizes SST/WAL files without changing the ZNS specification. The basic idea of Hi-ZNS is to put an SST/WAL file in a single zone. When an SST/WAL is invalidated, upper-level applications with HiZNS send the zone-reset command to a ZNS SSD device mindlessly without any space amplification and data migration. Specifically, Hi-ZNS allocates physical resources for the zone on demand and provides an infinite logical zone number to avoid wasting storage resources. The extensive evaluation demonstrates that Hi-ZNS substantially improves space efficiency and completely eliminates data migration. Compared with the existing ZenFS&RocksDB, HiZNS increases the maximum completed requests by up to 2.97x and improves the I/O performance by up to 7%.
引用
收藏
页码:1217 / 1226
页数:10
相关论文
共 23 条
  • [1] What You Can't Forget: Exploiting Parallelism for Zoned Namespaces
    Bae, Hanyeoreum
    Kim, Jiseon
    Kwon, Miryeong
    Jung, Myoungsoo
    [J]. PROCEEDINGS OF THE 2022 14TH ACM WORKSHOP ON HOT TOPICS IN STORAGE AND FILE SYSTEMS, HOTSTORAGE 2022, 2022, : 79 - 85
  • [2] Bjorling M, 2021, PROCEEDINGS OF THE 2021 USENIX ANNUAL TECHNICAL CONFERENCE, P689
  • [3] A Free-Space Adaptive Runtime Zone-Reset Algorithm for Enhanced ZNS Efficiency
    Byeon, Sungjin
    Ro, Joseph
    Jamil, Safdar
    Kang, Jeong-Uk
    Kim, Youngjae
    [J]. PROCEEDINGS OF THE 2023 15TH ACM WORKSHOP ON HOT TOPICS IN STORAGE AND FILE SYSTEMS, HOTSTORAGE 2023, 2023, : 109 - 115
  • [4] Choi G., 2020, P 12 USENIX WORKSH H, P1
  • [5] ApproxFTL: On the Performance and Lifetime Improvement of 3-D NAND Flash-Based SSDs
    Cui, Jinhua
    Zhang, Youtao
    Shi, Liang
    Xue, Chun Jason
    Wu, Weiguo
    Yang, Jun
    [J]. IEEE TRANSACTIONS ON COMPUTER-AIDED DESIGN OF INTEGRATED CIRCUITS AND SYSTEMS, 2018, 37 (10) : 1957 - 1970
  • [6] RocksDB: Evolution of Development Priorities in a Key-value Store Serving Large-scale Applications
    Dong, Siying
    Kryczka, Andrew
    Jin, Yanqin
    Stumm, Michael
    [J]. ACM TRANSACTIONS ON STORAGE, 2021, 17 (04)
  • [7] Han K, 2021, PROCEEDINGS OF THE 15TH USENIX SYMPOSIUM ON OPERATING SYSTEMS DESIGN AND IMPLEMENTATION (OSDI '21), P147
  • [8] Achieving Performance Isolation in Docker Environments with ZNS SSDs
    Han, Yejin
    Oh, Myunghoon
    Lee, Jaedong
    Yoo, Seehwan
    Kim, Bryan S.
    Choi, Jongmoo
    [J]. 2023 IEEE 12TH NON-VOLATILE MEMORY SYSTEMS AND APPLICATIONS SYMPOSIUM, NVMSA, 2023, : 25 - 31
  • [9] SplitZNS: Towards an Efficient LSM-Tree on Zoned Namespace SSDs
    Huang, Dong
    Feng, Dan
    Liu, Qiankun
    Ding, Bo
    Zhao, Wei
    Wei, Xueliang
    Tong, Wei
    [J]. ACM TRANSACTIONS ON ARCHITECTURE AND CODE OPTIMIZATION, 2023, 20 (03)
  • [10] Ji C., 2016, 8 USENIX WORKSH HOT