Btrfs RAID on zoned devices
summary
Btrfs gains a prototype for declustered parity that targets zoned devices by distributing extents so rebuilds read far fewer drives; data uses zoned append (kernel 5.12 behavior) while metadata remains serialized. A new stripe tree maps logical extents to per-device physical offsets and consumes ~16 bytes per extent (3+2 parity example uses ~80 bytes, still fits many nodes in a 4K sector). RAID1/RAID0 upgrades and background rebuilds are working, parity-algorithm plugins and full RAID5/6 erasure-code support remain future work.
AI-written summary · about
chapters (9)
- 0:00Introduction to Btrfs on zoned devices
- 1:04Btrfs basics and RAID support
- 2:11Zone block device support in Btrfs
- 4:17Lessons learned from RAID implementation
- 6:20Proposed RAID improvements
- 9:19Stripe tree design for RAID
- 13:37Metadata overhead and trade-offs
- 15:00Current status of RAID features
- 16:23Q&A on RAID and zoned devices
related talks
25mUsing Zoned Block Devices in Fedora
30mStRAID: Stripe-threaded Architecture for Parity-based RAIDs with Ultra-fast SSDs
30mOpenZFS and Linux
31mFrom Open-Channel SSDs to Zoned Namespaces
20mDeep Dive into Ceph Block Storage
24mMD cluster raid
33mEncryption for filesystems with advanced features new fscrypt functionality
14mAdvancing Data Integrity in Linux