The session's abstract was relatively mundane, "The Btrfs filesystem is quickly becoming a mature part of the Linux kernel. This talk will discuss the Btrfs roadmap and demonstrate some of the newest Btrfs features." And it was. It really wasn't too different from what was said at this year's Southern California Linux Expo.
Chris basically recapped the usual features talked about for Btrfs, iterated that the error-fixing Btrfs tool is still actively being worked on, and re-capped the big Btrfs file-system changes in the Linux 3.4 kernel.
- The file-system is jointly developed by many companies, not just Oracle. The other contributors range from Intel to SUSE.
- All data and meta-data written is via copy-on-write.
- CRCs are maintained for all meta-data and data.
- Btrfs provides for efficient writable snapshots, which allows for interesting features like system roll-backs and other great stuff.
- Multi-device support and built-in support for various RAID levels.
- Online resize and defragmentation support.
- Transparent compression via Zlib and LZO file-system compression.
- Efficient storage for small files.
- Solid-state drive optimizations (SSD) along with TRIM support.
The recent progress that Chris noted at this afternoon's San Francisco event included:
- Extensive performance and stability fixes.
- New and improving repair tool.
- Background scrubbing support.
- Automatic repair of corrupt blocks.
- RAID re-striping.
- Configurable meta-data block sizes.
- Improved I/O error handling infrastructure.
For those wondering about the Btrfs file-system performance compared to EXT4, XFS, and others, see Ubuntu 12.04 LTS - Benchmarking All The Linux File-Systems. From consumer hardware there's also Testing Out The Btrfs Mount Options On Linux 3.2. New benchmarks of hard drives and solid-state drives from the Linux 3.4 kernel will be forthcoming.