| 研究生: |
邱永陞 Chiu, Yung-Sheng |
|---|---|
| 論文名稱: |
設計與實作一個基於分區命名空間固態硬碟I/O緩衝方法之映射表適應性記憶體管理 Design and Implement an Adaptive Memory Management of Mapping Table based on an Approach to Buffering I/O for Zoned Namespace Solid-State Drive |
| 指導教授: |
侯廷偉
Hou, Ting-Wei |
| 學位類別: |
碩士 Master |
| 系所名稱: |
工學院 - 工程科學系 Department of Engineering Science |
| 論文出版年: | 2023 |
| 畢業學年度: | 111 |
| 語文別: | 中文 |
| 論文頁數: | 51 |
| 中文關鍵詞: | 固態硬碟 、位址映射 、分區命名空間固態硬碟 、映射表管理 |
| 外文關鍵詞: | solid-state drive, address mapping, Zoned Namespace SSD, mapping table management |
| 相關次數: | 點閱:52 下載:0 |
| 分享至: |
| 查詢本校圖書館目錄 查詢臺灣博碩士論文知識加值系統 勘誤回報 |
為了改善固態硬碟(Solid-State Drive, SSD)效能限制,開放通道固態硬碟(Open-Channel SSD, OCSSD)被提出,透過將快閃記憶體轉譯層(Flash Translation Layer, FTL)的大部分工作移至主機端執行,可有效改善SSD主控負載過高的問題。而分區命名空間固態硬碟(Zoned Namespace SSD, ZNS SSD)即是基於OCSSD架構的新型標準協議。ZNS SSD除了保有OCSSD的特點,進一步將邏輯位址劃分成不同的分區(zone),並且規定分區中的區塊(block)只能循序寫入,此作法大幅降低垃圾收集(garbage collection)所帶來的寫入放大問題。
然而,ZNS SSD循序寫入的限制可能會出現效能降級的問題。若在ZNS SSD之中增加緩衝區(buffer)架構做為緩衝並協同處理循序寫入的操作,雖可有效改善效能降級的問題,但會有記憶體使用量上升的缺點,且其映射表在初始化階段隨即一次性配置所有區塊(block)的映射資料於主記憶體中,可能造成記憶體浪費。
綜觀上述觀察,本論文針對映射表記憶體配置方式提出改善,設計一個兩層式映射表實現分層管理zone與block。模擬結果顯示在可接受的效能減損(約1.2%至8.5%)下,依據zone實際使用的狀態,適應性地動態釋放記憶體或動態配置記憶體,達成節省記憶體使用量,在最佳情況中可節省高達99%的映射表記憶體使用量,而在最差情況中的記憶體使用量相較原先單層式映射表僅多4072 bytes,並且同時保有使用緩衝區之優勢,比起無緩衝區架構的SPDK,效能仍保有15%至46%之提升。
To address performance degradation, Open-Channel Solid-State Drive (OCSSD) have been introduced, offloading partial FTL algorithms to the host to improve performance, alleviating SSD controller burdens. Zoned Namespace Solid-State Drive (ZNS SSD), an extension of OCSSD architecture, partition Logical Block Addresses (LBAs) into zones, enforcing sequential block writes. This approach significantly reduces garbage collection-induced write amplification, effectively enhancing SSD lifespan. However, sequential write constraint in ZNS SSD causes performance degradation in various scenarios.
Introducing buffer architecture within ZNS SSD to buffer and collaboratively process sequential writes mitigates performance degradation but escalates memory usage, and the one-level mapping table allocates all mapping data for all blocks in the mapping table into the main memory could potentially result in memory wastage.
This research proposes an improved memory allocation approach. The proposed two-level mapping table efficiently manages zones and blocks, adaptively releasing or allocating memory based on zone usage. This strategy significantly reduces memory consumption, achieving up to 99% reduction in mapping table memory usage in the best-case.
[1] Y. Li and K. N. Quader, "NAND flash memory: Challenges and opportunities", IEEE Computer, vol. 46, no. 8, pp. 23-29, Aug. 2013.
[2] C. Y. Lu, "Future prospects of NAND flash memory technology—The evolution from floating gate to charge trapping to 3D stacking", J. Nanosci. Nanotechnol., vol. 12, no. 10, pp. 7604-7618, 2012.
[3] S.S. Rizvi and Tae-Sun Chung, "Flash SSD vs HDD: High performance oriented modern embedded and multimedia storage systems", In 2nd International Conference on Computer Engineering and Technology (ICCET), Chengdu, China, 2010.
[4] K. Parat and A. Goda, "Scaling Trends in NAND Flash", 2018 IEEE International Electron Devices Meeting (IEDM), pp. 2.1.1-2.1.4, San Francisco, CA, USA, 2018.
[5] C. M. Compagnoni, A. Goda, A. S. Spinelli, P. Feeley, A. L. Lacaita and A. Visconti, "Reviewing the evolution of the NAND flash technology", Proc. IEEE, vol. 105, no. 9, pp. 1609-1633, Sep. 2017.
[6] T.-S. Chung, D.-J. Park, S. Park, D.-H. Lee, S.-W. Lee and H.-J. Song, "A survey of flash translation layer", J. Syst. Archit., vol. 55, no. 5, pp. 332-343, May/Jun. 2009.
[7] J. Kim, K. Lim, Y. Jung, S. Lee, C. Min and S. H. Noh, "Alleviating garbage collection interference through spatial separation in all flash arrays", Proc. USENIX Annu. Techn. Conf., pp. 799-812, Renton, WA, USA, 2019.
[8] I. Picoli, N. Hedam, P. Bonnet and P. Tzn, "Open-Channel SSD (what is it good for)", Conference on Innovative Data Systems Research. Amsterdam, Netherlands, 2020.
[9] M. Bjørling, "From open-channel SSDs to zoned namespaces", Proc. Linux Storage Filesyst. Conf. (Vault), pp. 20, BOSTON, MA, USA, Feb. 2019.
[10] T. Stavrinos, D. S. Berger, E. Katz-Bassett and W. Lloyd, "Don’t be a blockhead: zoned namespaces make work on conventional SSDs obsolete", Workshop on Hot Topics in Operating Systems (HotOS), pp. 144-151, June, 1-3, Michigan, USA, 2021.
[11] M. Bjørling, A. Aghayev, H. Holmberg, A. Ramesh, D. L. Moal, G. R. Ganger, et al., "ZNS: avoiding the block interface tax for flash-based SSDs", In Proceedings of the 2021 USENIX Annual Technical Conference (USENIX ATC'21), pp. 689-703, 2021.
[12] Western Digital Corporation, "Zoned Storage," [Online]. Available: https://zonedstorage.io/. [Accessed 30 July 2023].
[13] Youngjae Lee, Jeeyoon Jung, and Dongkun Shin., "Buffered i/o support for zoned namespace SSD", In 2021 IEEE International Conference on Consumer Electronics-Asia (ICCE-Asia), pages 1–4, 2021.
[14] W.-S. Huang, An approach to buffering I/O in host memory for Zoned Namespace solid-state drive, M.S. Thesis, Dept. of Engineering Science, National Cheng Kung University, Tainan, Taiwan, 2023.
[15] C.-H. Wu, Design and implement a face recognition system combining embedded device and mobile phone application, M.S. Thesis, Dept. of Engineering Science, National Cheng Kung University, Tainan, Taiwan, 2022.
[16] J.-L. Lin, An I/O tracing tool for Zoned Namespace solid-state drive based on SPDK,M.S. Thesis, Dept. of Engineering Science, National Cheng Kung University, Tainan, Taiwan, 2023.