Data Storage in Cloud Native AI (CNCF Technical Community Group Meeting)

Data Storage in Cloud Native AI

Part of the CNCF Artificial Intelligence Technical Community Group Meeting Series.

Xing Yang, Kubernetes Storage SIG co-chair and co-author of the CNCF TAG Infrastructure whitepaper Data Storage in Cloud Native AI, will walk through the paper's key findings on how storage architecture must evolve to keep up with modern AI workloads.

If you're running AI on Kubernetes, your storage layer is probably a bottleneck. This whitepaper is a comprehensive guide the CNCF community has produced on the topic.

What You'll Learn

• Why training, inference, and AI agents each need fundamentally different storage strategies and what breaks when you get it wrong
• How checkpointing failures and GPU idle time from slow data pipelines can silently burn your compute budget
• Practical patterns: caching hierarchies, vector databases for RAG, KV caching for generative AI, and model loading optimization for cold starts
• Where object storage, file systems, and block storage each fit and where they don't
• How emerging architectures like Data Mesh and Data Fabric change the picture for AI/ML teams
• The hardware/software co-design frontier: CXL, domain-specific architectures, and what comes after Moore's Law

Whitepaper: Data Storage in Cloud Native AI

Joining Details

Please use your LF username and password to avoid challenges with the joining URL. Alternatively, use the Zoom login credentials to join the session directly.

Join from PC, Mac, iPad, or Android:

• Join link: Click here to join
• Meeting ID: 96058530101
• Meeting Passcode: 290975