일 | 월 | 화 | 수 | 목 | 금 | 토 |
---|---|---|---|---|---|---|
1 | 2 | 3 | 4 | 5 | 6 | 7 |
8 | 9 | 10 | 11 | 12 | 13 | 14 |
15 | 16 | 17 | 18 | 19 | 20 | 21 |
22 | 23 | 24 | 25 | 26 | 27 | 28 |
29 | 30 |
- Flash Memory
- FTL
- Machine Learning
- hardware
- rocksdb
- 시스템 소프트웨어
- linux
- 커널 프로그래밍
- core dumped
- overflow
- 키워드
- 시스템 프로그래밍
- Cache
- Intel
- framework
- USENIX
- Operating System
- Samsung
- kernel
- Git
- 포트 번호 변경
- deep learning
- github
- ssd
- performance
- memory
- software
- storage system
- Today
- Total
Happy to visit my research note ^^
(관심 논문) Multi-view Feature-based SSD Failure Prediction:What, When, and Why 본문
(관심 논문) Multi-view Feature-based SSD Failure Prediction:What, When, and Why
Liam Lim 2023. 3. 21. 20:21Yuqi Zhang† , Wenwen Hao† , Ben Niu‡ , Kangkang Liu‡ , Shuyang Wang† , Na Liu† , Xing He† , Yongwong Gwon∗ , Chankyu Koh∗ †Samsung R&D Institute China Xi’an, Samsung Electronics ‡Tencent ∗Samsung Electronics
February 21–23, 2023 • Santa Clara, CA, USA
USENIX Association
21st USENIX Conference on File and Storage Technologies
Abstract
SSD (Solid-State Drive)는 large-scale data centers에서 중요한 역할을 한다. SSD`s failures는 storage system의 stability에 영향을 미치고 additional maintenance overhead의 원인이 된다. 미리 SSD`s failures를 예측하고 처리하기 위해, 본 논문은 multi-view and multi-task random forest (MVTRF) scheme을 제안한다. MVTRF는 SSD의 long-term and short-term monitoring data로 부터 mujlti-view features를 기반으로 failures를 예측한다. 특히, multi-task learning은 동일한 모델을 통해 고장 유형과 언제 고장이 발생할 것인지 동시에 예측한다. 또한, 저자들은 MVTRF의 key decisions를 추출하여 고장이 발생하는 이유를 분석한다. 이러한 details of failure는 SSD`s failures를 verifying and handling하는 데 유용하다. Proposed MVTRF는 data centers의 large-scale real data에서 평가된다. 실험 결과, MVTRF는 existing scheme과 비교해서 prediction accuracy가 높고 precision이 평균 46.1%, recall (재현율)이 57.4% 향상되었음을 보여준다. 결과는 또한 failure type and time prediction and failure cause identification에 효과적인 것으로 입증되었고, 이는 failure handling의 efficiency를 향상시키는 데 도움이 된다.