Famfs Update: Status, DAX Challenges & Use Cases

on YouTube ↗15m

" summary

Famfs formats devdax-backed memory as a scale-out POSIX file system and aims for upstream inclusion next year, while navigating a recent DAX refactor that broke FS-on-devdax behavior and required a custom kernel branch to restore it. The implementation maps file data via FS DAX, keeps metadata append-only, and limits creators to one node today. Measured use: a 4.6 TB graph workload on four nodes drops from ~7.5 hours to ~15 minutes. first file system fully hosted on devdax upstream goal mid-next-year DAX layout_final break required API adjustments

AI-written summary · about

" chapters (12)

1. 0:00Introduction to Famfs
2. 1:03What is fabric attached memory
3. 2:28Core insight behind Famfs
4. 3:12DAX challenges and upstreaming
5. 5:19DAX layout final API
6. 6:05Famfs use cases overview
7. 7:31CXL memory latency comparison
8. 8:35Graph analysis workload demo
9. 10:42Roadmap and upstreaming plan
10. 11:02Scale up interconnects discussion
11. 12:29Hardware cache coherency
12. 13:55Wrap-up and Q&A