プレプリント / バージョン1

I/O Scheduler using splitting mechanism for improving Real-Time performance in eMMC

##article.authors##

  • Shinnosuke Koshiba Graduate School of Informatics, Nagoya University
  • Yutaka Matsubara Graduate School of Informatics, Nagoya University
  • Hiroaki Takada Graduate School of Informatics, Nagoya University

DOI:

https://doi.org/10.51094/jxiv.1062

キーワード:

I/O scheduler、 eMMC

抄録

Embedded MultiMedia Card (eMMC) is widely used in many real-time systems, such as in automotive applications. In embedded systems, it is common for applications to share storage resources, which creates a demand for protecting real-time processes from interference by other processes. Since eMMC lacks internal command queu- ing, the I/O scheduler becomes a critical component in ensuring performance. Under these conditions, priority-based schedulers for real-time systems have been developed, successfully protecting real-time processes. However, our ex- periments revealed that while these schedulers can handle the bandwidth load caused by a large number of small-sized I/O requests, they fail to manage the load caused by a small number of large-sized I/O requests. In such cases, tail latency of real-time process significantly worsens to levels comparable to systems without a scheduler. In this research, we propose an I/O scheduler equipped with an I/O splitting mechanism that converts a small number of large-sized I/O operations into a large number of small-sized I/O operations. Experiments demonstrated that, compared to ex- isting priority-based schedulers, this approach successfully reduced 99th percentile latency, achieving up to a 5-fold reduction in read latency and up to a 3-fold reduction in write latency.

利益相反に関する開示

There are no conflicts of interest for any of the authors.

ダウンロード *前日までの集計結果を表示します

ダウンロード実績データは、公開の翌日以降に作成されます。

引用文献

B. Jun and D. Shin:Workload-aware budget compensation scheduling for NVMe solid state drives, in Proceedings of the 2015 Non-Volatile Memory System and Applications Symposium (NVMSA), Hong Kong, China: IEEE, Aug. 2015, pp. 19–24.

J. Kim, S. Seo, D. Jung, J.-S. Kim, and J. Huh.:Parameter- aware I/O management for solid state disks (SSDs), IEEE Transactions on Computers, vol. 61, no. 5, pp. 636–649, May 2012.

A. Tavakkol, M. Sadrosadati, S. Ghose, J. S. Kim, Y. Luo, Y. Wang, N. M. Ghiasi, L. Orosa, J. Go ́mez-Luna, and O. Mutlu.:FLIN: Enabling fairness and enhancing perfor- mance in modern NVMe solid state drives, in Proceedings of the 45th Annual International Symposium on Computer Architecture (ISCA), Los Angeles, CA, USA: IEEE, Jun. 2018, pp. 397–410.

S. Park and K. Shen.:FIOS: A fair, efficient flash I/O scheduler, in Proceedings of the 10th USENIX Conference on File and Storage Technologies (FAST), San Jose, CA, USA: USENIX, Feb. 2012, pp. 155–170.

Jung, M.; Choi, W.; Srikantaiah, S.; Yoo, J.; Kandemir, M.T.:HIOS: A host interface I/O scheduler for Solid State Disks, In Proceedings of the Annual International Sym- posium on Computer Architecture (ISCA), Minneapolis, MN, USA, 14–18 June 2014.

Till Miemietz, Hannes Weisbach, Michael Roitzsch, and Hermann Ha ̈rtig.:K2: Work-Constraining Scheduling of NVMe-Attached Storage, In Proceedings of 2019 IEEE Real-Time Systems Symposium (RTSS ’19), pages 56–68, 2019.

F. Blagojevic, C. Guyot, Q. Wang, T. Tsai, R. Mateescu and Z. Bandic.:Priority IO scheduling in the cloud, Proc. USENIX Conf. Hot Topics Cloud Comput., pp. 1-6, 2013- Jun.

Kyusik Kim, Seungkyu Hong, and Taeseok Kim.:Supporting the priorities in the multi-queue block I/O layer for NVMe SSDs, J. Semiconduct. Technol. Sci. 20 (02 2020), 55–62.

MQ-Deadline Implementation,https://elixir. bootlin.com/linux/latest/source/block/mq- deadline.c,Accessed: 2024-10-02.

F. Wu, H. Xi, and J. Li.:Linux readahead: less tricks for more, in Proceedings of the Linux Symposium, 2007.

Queue sysfs,https://www.kernel.org/doc/ Documentation/block/queue-sysfs.txt, Accessed: 2024-10-02.

Caeden Whitaker, Sidharth Sundar, Bryan Harris, and Ni- hat Altiparmak.:Do we still need io schedulers for low- latency disks?, In 10th USENIX Workshop on Hot Topics in Storage and File Systems (HotStorage 23), HotStorage ’23, 2023.

Ren, Zebin, Krijn Doekemeijer, Nick Tehrany, and Ani- mesh Trivedi.:BFQ, Multiqueue-Deadline, or Kyber? Performance Characterization of Linux Storage Schedulers in the NVMe Era. ,In Proceedings of the 15th ACM/SPEC International Conference on Performance Engineering, pp. 154-165. 2024.

O. Sandoval, blk-mq: Introduce Kyber mul- tiqueue I/O scheduler, Commit message for 10 the Linux kernel, 2017. [Online]. https://patchwork.kernel.org/patch/9672023 2024-10-02.

D. Zhou, W. Pan, W. Wang and T. Xie.:I/O characteristics of smartphone applications and their implications for eMMC design, Proc. of IEEE IISWC, 2015.

Missimer, Katherine, and Richard West.:Partitioned real- time NAND flash storage.,2018 IEEE Real-Time Systems Symposium (RTSS). IEEE, 2018.

Qin, Z.; Wang, Y.; Liu, D.; Shao, Z.:Real-Time Flash Translation Layer for NAND Flash Memory Storage Sys- tems, In Proceedings of the IEEE Real-Time Technol- ogy and Applications Symposium (RTAS), Beijing, China, 16–19 April 2012.

Lee, Jongmin, et al.:Real-time flash memory storage with Janus-FTL, Proceedings of the 27th Annual ACM Sympo- sium on Applied Computing. 2012.

Chang, L.P.; Kuo, T.W.; Lo, S.:Real-time garbage collec- tion for flash-memory storage systems of real-time embed- ded systems, ACM Trans. Embed. Comput. Syst. 2004, 3, 661–863.

Choudhuri, S.; Givargis, T.D.:Deterministic service guar- antees for nand flash using partial block cleaning, In Pro- ceedings of the 6th IEEE/ACM/IFIP International Confer- ence on Hardware/Software Codesign and System Synthe- sis, Atlanta, GA, USA, 19–24 October 2008.

Zhang, Q.; Li, X.; Wang, L.; Zhang, T.; Wang, Y.; Shao, Z.:Optimizing Deterministic Garbage Collection in NAND Flash Storage Systems, In Proceedings of the IEEE Real- Time Technology and Applications Symposium (RTAS), Seattle, WA, USA, 13–16 April 2015.

Missimer, K.; West, R.:Partitioned Real-Time NAND Flash Storage, In Proceedings of the IEEE Real-Time Sys- tems Symposium (RTSS), Nashville, TN, USA, 11–14 De- cember 2018.

eBPF Documentation. https://ebpf.io/ what-is-ebpf/. Accessed: 2024-10-02.

bpftrace. https://bpftrace.org/.Accessed: 2024-10- 02.

ダウンロード

公開済


投稿日時: 2025-01-27 09:59:22 UTC

公開日時: 2025-01-29 08:46:00 UTC
研究分野
情報科学