Textbook of Linux

XFS is a 64-bit, high-performance journalling filesystem originally developed by Silicon Graphics for IRIX in 1993 and ported to Linux in 2001. It excels at large files, large filesystems (up to 8 EiB), and highly parallel workloads, thanks to its use of B+ trees throughout and allocation groups that let multiple CPUs perform metadata operations independently. XFS is the default filesystem on Red Hat Enterprise Linux and CentOS Stream.

Distinguishing features include delayed allocation (blocks are chosen just before being written, reducing fragmentation), online resizing (growing only; XFS cannot shrink), online defragmentation with xfs_fsr, and project quotas. Metadata is CRC-checksummed by default since about 2013, catching on-disk corruption. The journal contains only metadata, not data, so crash recovery is fast but does not protect against torn writes in file content—an issue modern workloads handle with O_DIRECT or application-level checksums.

sudo mkfs.xfs /dev/sdb1
sudo xfs_info /mnt
sudo xfs_growfs /mnt            # grow after extending the block device
sudo xfs_repair /dev/sdb1       # offline repair

XFS is a reasonable choice for large databases, big-data storage, and file servers. It lacks the snapshots and checksummed data of btrfs/zfs, but for raw throughput and stability on enterprise hardware, few Linux filesystems match it.