Marc Merlin, a Linux admin at Google for more than one decade, is presenting on Thursday at LinuxCon Chicago about Btrfs. His slides are already available for those that can't make it to the windy city or are looking for an overview of what he'll be discussing.
Here's the highlights to Merlin's presentation:
- Marc has been running Btrfs on his laptop for the past two years and on other Linux systems more recently.
- Btrfs is recommended for its copy-on-write (COW) capabilities, snapshots are built into the file-system and offer good performance, meta-data is redundant and check-summed as well as the data, RAID 0/1/5/6 are built into the file-system, no need for multiple partitions or LVM volumes, built-in file compression, online background file-system scrubbing, block-level file-system diff backups, and support with btrfs-convert for converting an EXT3 file-system to Btrfs.
- Against ZFS, the former Sun file-system is more mature and offers slightly more features than Btrfs but its license is incompatible with the mainline Linux kernel and the file-system is very RAM hungry. It's unlikely Oracle will ever relicense ZFS from the CDDL to GPLv2 and offer a native port for the mainline Linux kernel.
- Oracle supports Btrfs in its commercial distribution, "basic Btrfs" is considered mostly stable, RAID 0/1 should be in good shape, Btrfsck repair tool is still incomplete, data de-duplication support is experimental, and file-system encryption isn't yet natively implemented.
- Many companies contribute to Btrfs from Facebook to Fusion-IO to Intel to Red Hat to SUSE.
- In terms of Btrfs production readiness from Marc's perspective, "If you pick the right btrfs release/kernel (you can let Oracle or Suse do this for you), you can look at Btrfs for production use." Additionally, "In a nutshell, while Btrfs is still experimental, it is usable in production environment for its core features."