Longtime Linux developer Christoph Hellwig has announced the first public release of XRAID, Phoronix reports: a declustered parity RAID implementation integrated into the XFS file system, aimed especially at zoned XFS setups such as SMR hard drives. It also works on conventional disks and SSDs, though SSD performance hasn’t been the main focus. Hellwig says writes can be up to 20 times faster than mdraid, even without that tuning.
The design uses XFS’s zoned allocator, which keeps file data on the realtime subvolume and metadata elsewhere, so XRAID only has to do parity RAID on the file data. Metadata redundancy is left to existing stacked drivers such as MD RAID mirroring, which suits small objects that get updated in place. Declustered RAID spreads data and parity across a pool larger than a single stripe, so rebuilds can use spare capacity on every disk, and the rebuild order can favor the zones that need it most.
This is a public code drop, not a claim that your distro kernel already ships it. Hellwig’s announcement is on the Linux kernel mailing list if you want the technical write-up.
Review before you install: read the release notes yourself, back up your config, and try new versions somewhere you can afford to break first.
My take
RAID inside the file system, built for zoned devices, is the kind of unglamorous infrastructure work that matters if you have a pile of big disks. I’ll wait for wider review before I’d trust those “20 times faster” numbers on a machine I care about. If you experiment, do it on a box you can afford to rebuild.
God Bless, and Use the Source!
