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 | 31 |
- overflow
- github
- memory
- deep learning
- software
- USENIX
- Cache
- 시스템 소프트웨어
- linux
- 키워드
- hardware
- performance
- Operating System
- storage system
- kernel
- 커널 프로그래밍
- Git
- Samsung
- core dumped
- Flash Memory
- Machine Learning
- ssd
- framework
- Intel
- 시스템 프로그래밍
- FTL
- 포트 번호 변경
- rocksdb
- Today
- Total
Happy to visit my research note ^^
(관심 논문) More Than Capacity: Performance-orientedEvolution of Pangu in Alibaba 본문
(관심 논문) More Than Capacity: Performance-orientedEvolution of Pangu in Alibaba
Liam Lim 2023. 3. 21. 19:03Qiang Li⋄ , Qiao Xiang† , Yuxin Wang†⋄ , Haohao Song†⋄ , Ridi Wen†⋄ , Wenhui Yao⋄ , Yuanyuan Dong⋄ , Shuqi Zhao⋄ , Shuo Huang⋄ , Zhaosheng Zhu⋄ , Huayong Wang⋄ , Shanyang Liu⋄ , Lulu Chen⋄ , Zhiwu Wu⋄ , Haonan Qiu⋄ , Derui Liu⋄ , Gexiao Tian⋄ , Chao Han⋄ , Shaozong Liu⋄ , Yaohui Wu⋄ , Zicheng Luo⋄ , Yuchao Shao⋄ , Junping Wu⋄ , Zheng Cao⋄ , Zhongjie Wu⋄ , Jiaji Zhu⋄ , Jinbo Wu⋄ , Jiwu Shu† , Jiesheng Wu⋄ , ⋄Alibaba Group, †Xiamen University
February 21–23, 2023 • Santa Clara, CA, USA
USENIX Association
21st USENIX Conference on File and Storage Technologies
Abstract
본 논문은 Pangu storage system이 hardware technologies and business model에 따라 지속적으로 발전하여 100 ㎲ level의 I/O latency로 high-performance, reliable storage services를 제공하는 방법을 제시한다. Pangu의 발전은 두 가지 단계로 이루어졌다.
- 첫 번째 단계에서는 Pangu는 SSD storage와 RDMA (Remote Direct Memory Access) network technologies의 출현을 수용해서 file system을 혁신하고 user-space storage operating system를 설계하여 I/O latency를 크게 줄이면서 high-throughput과 IOPS를 제공했다.
- 두 번째 단계에서는, Pangu는 volume-oriented storage provider (공급 업체)에서 performance-oriented provider로 진화했다.
이 business model의 변화에 적응하기 위해, Pangu는 SSDSSD 용량이 훨씬 더 재밌고 RDMA bandwidth가 25 Gbps에서 100 Gbps로 업그레이드된 storage server의 infrastructure를 도입했다. Traffic amplification reduction, remote direct cache access and CPU computation offloading와 같은 일련의 핵심 설계를 도입하여 hardware upgrade에 따른 performance improvement를 완전히 수확할 수 있도록 보장했다. Technology innovations 외에도, operating experience를 공유하고 이를 통해 얻은 중요한 교훈에 대해 논의했다.