CubeFS in Action: Empowering Users Through Case Studies
Talkat KubeCon + CloudNativeCon Europe 2025 · on Friday, 2025-04-04· at 11:00· Level 3 | ICC Capital Suite 14-16·
summary
CubeFS is a CNCF‑graduated, cloud native filesystem with S3 and POSIX interfaces, two storage engines (multi‑replica and erasure coding), and Raft‑based strong consistency. It powers OPPO AI pipelines with a distributed cache that shards files into 1 MB segments and benchmarks each 4 TB cache node at 50 GB/s throughput, uses lifecycle tiering after observing 80% of AI data cold within three months, and replaces FUSE with an SDK to meet P99 <1 ms latency for key‑value workloads.
AI-written summary · about
chapters (9)
- 0:00Introduction to CubeFS
- 1:38Architecture and features
- 5:07AI storage challenges
- 7:24AI data management
- 14:06Distributed cache solutions
- 18:44Intelligent data tiering
- 20:36Compute-storage separation
- 24:36SDK and optimizations
- 26:21Future prospects
related talks
33mThe Evolution and Innovation of CubeFS
42mProduction FS: Adapt or Die
31mScaling Autonomous-Driving R&D: A 100 PB, 100 Billion-File Open-Source Elastic File...
14mFalconFS: Distributed File System for Large-Scale Deep Learning Pipeline
09mShared cloud file storage, uses cases, and the latest Amazon EFS innovations
47mDesign Decisions Behind Cube Store - a Blazingly Fast Open-Source Rollup Storage Layer
36mStorage Wars
27mModernize your applications with serverless, containers, and Amazon EFS