Linux kernel developer Christoph Hellwig has announced the initial public release of XRAID, a new declustered parity RAID implementation integrated with the XFS filesystem.
According to Hellwig, XRAID has been in development for about 18 months, with two main goals: to enable parity RAID on zoned devices and deliver high performance without the data integrity problems of some traditional RAID approaches.
Although designed primarily for zoned storage, XRAID also works with conventional hard drives and SSDs. Early benchmarks show write performance up to 20 times faster than Linux’s existing mdraid implementation, despite SSD optimization not being a priority.
One of XRAID’s main architectural differences is its integration with XFS. It builds on the filesystem’s zoned allocator, which separates file data from metadata using the XFS realtime subvolume. File data is placed in sequential-write zones, while metadata is stored on conventional devices or zones.
This separation allows XRAID to apply parity protection specifically to file data, where it is most efficient. Metadata redundancy is instead handled by existing Linux mechanisms, such as MD RAID mirroring, which is better suited to smaller objects that need frequent in-place updates.
Another important feature is XRAID’s declustered architecture. Unlike conventional RAID configurations where each stripe spans a fixed set of disks, declustered RAID distributes data and parity throughout a larger storage pool.
The approach can greatly reduce rebuild times after disk failures. XRAID reconstructs data across multiple devices with available capacity and automatically prioritizes zones that lost multiple levels of redundancy. It also rebuilds data zones before parity zones to reduce the performance impact of degraded reads.
The initial implementation supports configurations with zero, one, or two parity blocks. The developers have also experimented with the SnapRAID parity library to support three or more parity blocks. That functionality is not yet included, although the code has been prepared for future expansion.
On the management side, a dedicated branch of xfsprogs introduces filesystem creation support and a new xraid utility for managing devices and inspecting XRAID filesystems. The developers have also extended Zonar, a zoned filesystem visualization tool, to display how logical zones map to physical storage devices.
However, XRAID remains experimental. Hellwig explicitly states that the implementation is not yet ready for submission to the mainline Linux kernel. The public release is intended to collect feedback and encourage broader testing before upstream integration is considered.
For those interested in the technical details, the developers have published the XRAID kernel source code, along with corresponding xfsprogs and filesystem testing repositories. Additional architectural information and benchmark results are available in Hellwig’s FOSSY 2026 presentation.
