CSAL: the Next-Gen Local Disks for the Cloud
Yanbo Zhou, Erci Xu, Li Zhang, Kapil Karkra, Mariusz Barczak, Wayne Gao, Wojciech Malikowski, Mateusz Kozlowski, Lukasz Lasek, Ruiming Lu, Feng Yang, Lilong Huang
Abstract
Cloud local disks are attractive for their affordable price and high performance. The recent advancement in CPUs motivates cloud vendors to further multiplex the computing resources to serve more users. Unfortunately, such proposals are constrained by the limited offerings of cloud local disks per server as the underlying storage devices are either large but slow (e.g., HDDs) or fast yet small (e.g., NVMe SSDs).
In this paper, we explore the possibility of leveraging highcapacity QLC-based SSDs for cloud local disks. However, the three preliminary unsuccessful attempts indicate that QLC SSDs cannot simply work as drop-in replacement. The root cause is the two levels of write amplification caused by device-level address mapping with Indirection Unit and NAND-level garbage collection.
With these lessons learned, we propose CSAL, the nextgen local disks in Alibaba Cloud. CSAL includes a highperformance SSD as write buffers and a large-capacity QLC SSD for persistence. With a two-level Logical to Physical (L2P) address mapping table, CSAL achieves fine-grained (4KB) data accessing and significantly alleviates the two levels of write amplification. Results show that CSAL always prevails with superior performance and can achieve up to 2.22×, 1.82×, and 2.03× speedups against the second-best peers in micro, application, and deployment benchmarking, respectively. As of now, we have deployed CSAL on thousands of servers and made CSAL open-source to the public.
Ask about this paper
Your agent reads all of it.
Lune indexed this paper to the last equation, along with the top-tier papers that cite it. Ask a question and the answer quotes them.
Your agent calls
Luneget_paper_fulltext
Free to start. No credit card required.
Terminal
Install the CLIlune papers fulltext 40fb13de-e6d1-4b9f-8e6e-7d92a3a31aefCited by top-tier papers6
- Reviving In-Storage Hardware Compression on ZNS SSDs through Host-SSD CollaborationYingjia Wang, Tao Lu, Yuhong Liang, Xiang Chen et al.HPCA 2025 · 7 citations
- Z-LFS: A Zoned Namespace-tailored Log-structured File System for Commodity Small-zone ZNS SSDsInhwi Hwang, Sangjin Lee, Sunggon Kim, Hyeonsang Eom et al.USENIX ATC 2025 · 7 citations
- Here, There and Everywhere: The Past, the Present and the Future of Local Storage in CloudLeping Yang, Yanbo Zhou, Gong Zeng, Li Zhang et al.FAST 2026 · 2 citations
- RosenBridge: A Framework for Enabling Express I/O Paths Across the Virtualization BoundaryShi Qiu, Li Wang, Jianqin Yan, Ruofan Xiong et al.FAST 2026 · 1 citation
- Rearchitecting Buffered I/O in the Era of High-Bandwidth SSDsYekang Zhan, Tianze Wang, Zheng Peng, Haichuan Hu et al.FAST 2026
Builds on5
- ZNS: Avoiding the Block Interface Tax for Flash-based SSDsMatias Bjørling, Abutalib Aghayev, Hans Holmberg, Aravind Ramesh et al.USENIX ATC 2021 · 221 citations
- Separating Data via Block Invalidation Time Inference for Write Amplification Reduction in Log-Structured StorageQiuping Wang, Jinhong Li, Patrick P. C. Lee, Tao Ouyang et al.FAST 2022 · 56 citations
- Improving the Reliability of Next Generation SSDs using WOM-v CodesShehbaz Jaffer, Kaveh Mahdaviani, Bianca SchroederFAST 2022 · 26 citations
- Spool: Reliable Virtualized NVMe Storage Pool in Public Cloud InfrastructureShuai Xue, Shang Zhao, Quan Chen, Gang Deng et al.USENIX ATC 2020 · 17 citations
- IPLFS: Log-Structured File System without Garbage CollectionJuwon Kim, Minsu Kim, Muhammad Danish Tehseen, Joontaek Oh et al.USENIX ATC 2022
Related papers
- Burstable Cloud Block Storage with Data Processing UnitsJunyi Shu, Kun Qian, Ennan Zhai, Xuanzhe Liu et al.OSDI 2024 · 17 citations
- POLARDB Meets Computational Storage: Efficiently Support Analytical Workloads in Cloud-Native Relational DatabaseWei Cao, Yang Liu, Zhushi Cheng, Ning Zheng et al.FAST 2020 · 140 citations
- The Unwritten Contract of Cloud-based Elastic Solid-State DrivesYingjia Wang, Ming-Chang YangDAC 2025 · 2 citations
- From luna to solar: the evolutions of the compute-to-storage networks in Alibaba cloudRui Miao, Lingjun Zhu, Shu Ma, Kun Qian et al.SIGCOMM 2022 · 79 citations
- StreamCSD: SSD-Autonomous Stream Management via In-Storage Content LearningWenjie Li, Xiang Chen, Yelin Shan, Jiapin Wang et al.DAC 2025
