DOI QR코드

DOI QR Code

논리 볼륨 매니저를 이용한 파일 우선순위 기반의 하이브리드 저장장치 관리 시스템

Priority-Based Hybrid File Storage Management System Using Logical Volume Manager

  • Choi, Hoonha (Department of Computer Engineering, Sejong University) ;
  • Kim, Hyeunjee (Department of Computer Engineering, Sejong University) ;
  • No, Jaechun (Department of Computer Engineering, Sejong University)
  • 투고 : 2016.08.20
  • 심사 : 2016.11.22
  • 발행 : 2016.12.25

초록

최근 고성능의 SSD(Solid State Drive)가 등장하면서 단일노드의 입출력 성능이 대폭 상향되었다. 이에 SSD를 기반으로 하는 차세대 저장장치 플랫폼이 주목을 받게 되었고, 고속 연산이 필요한 서버 또는 데이터 센터 등에서 SSD 기반 저장장치를 구축하는 시도가 증가하고 있다. 그러나 SSD는 단위용량당 비용이 고가이기 때문에, SSD 기반 저장장치를 구성하기에는 아직 어려움이 있다. 따라서, 본 논문은 저가와 큰 용량이 장점인 HDD(Hard Disk Drive)와 SSD가 통합된 저장장치에서 파일을 관리하는 소프트웨어 HyPLVM(Hybrid Priority Logical Volume Manager)을 소개한다. HyPLVM은 사용자가 접근하는 파일, 디렉터리를 분석하여 파일에 우선순위를 부여하고, 이 우선순위 값에 따라 높으면 SSD에, 낮으면 HDD에 저장되도록 관리한다. 이로써, 접근빈도가 많은 파일에 한에서 SSD로 구성된 저장장치와 버금가는 입출력 성능을 산출하면서 저장장치 구축비용을 절감한다.

Recently, the I/O performance of a single node is rapidly improving due to the advent of high-performance SSD. As a result, the next-generation storage platform based on SSD has received a great deal of attention and such storage platforms are increasingly adopted to commodity servers or data centers that look for the high-bandwidth computation and I/O. However, building all SSD-based storage platform may not be cost-effective because the price per storage capacity is very high as compared to that of HDD. In this paper. we propose a hybrid file management solution, called HyPLVM(Hybrid Priority Logical Volume Manager), which combines the strength of SSD with the desirable aspects of low-price, high-storage capacity HDD. HyPLVM prioritizes the files and directories to be accessed by users, in order to determine the target storage device (SSD/HDD) in which files are allocated, while mitigating the cost of building storage platforms.

키워드

참고문헌

  1. Jeong-Su Park, Yu-Mi Bae, Sung-Jae Jung, "Technical analysis of Cloud Storage for Cloud Computing", KIICE, Vol. 17, no. 5, pp. 1129-1137, May 2013.
  2. Yaffs. Available on July 2011 from "http://www.yaffs.net/"
  3. D. Woodhouse. Jffs: The journalling flash file system. In The Ottawa Linux Symposium, RedHat Inc, 2001.
  4. TAILWINDSTORAGE. Extreme 3804, "http://tailwindstorage.com/products/"
  5. Fusion-IO, "iodrive octal data sheet, http://www.fusionio.com/data-sheets/iodrive-octal-data-sheet/"
  6. D. G. Andersen and S. Swanson, "Rethinking flash in the data center", IEEE Micro, Vol. 30, no. 4, pp.52-54, Jul. 2010. https://doi.org/10.1109/MM.2010.71
  7. F. Douglis, R. Caceres, F. Kaashoek, K. Li, B. Marsh, J. Tauber, "Storage alternatives for mobile computers", Proceedings of the 1st USENIX conference on Operating Systems Design and Implementation, 1994.
  8. Jin-Yong Ha, Jin-Soo Kim, "Hext4: A Filesystem for SSD-HDD Hybrid Storage", KIISE, pp. 1403-1405, Jun, 2014.
  9. Sanghyun Yoo, Kyung Tae Kim, Hee Yong Youn, "Block Replacement Scheme based on Reuse Interval for Hybrid SSD System", KSII, Vol 16, no. 5, pp. 19-27, Oct, 2015.
  10. Amplicon, "Benefits of SSD vs. HDD, https://www.amplicon.com/docs/white-papers/SSD-vs-HDD-white-paper.pdf"
  11. AJ Lewis, "LVM HOWTO, http://www.sistina.com"
  12. David Teigland, "Volume Manager in Linux", Sistina Software, Inc., 2001.
  13. redhat, "The Linux Logical Volume Manager, http://www.redhat.com/magazine/009jul05/features/lvm2/"
  14. Siwoo Byun, "A Hetero-Mirroring Scheme to Improve I/O Performance of High-Speed Hybrid Storage", KAIS, Vol. 11, No. 12, pp. 4997-5006, 2010.
  15. Sanghyun Yoo, Hee Yong Youn, "Cache replacement algorithm based on time and hit ratio in NAND flash memory", KSII, pp. 231-232, Oct 2014.