Linux Filesystems: The Complete Comparison Guide — ext4 vs XFS vs Btrfs vs ZFS

"How should I format this?" The answer to that question changes completely with your purpose. ext4 is the king of stability, XFS the master of large-volume handling, Btrfs the wizard of snapshots, and ZFS the enterprise endgame.

1. What Is a Filesystem

A filesystem is the set of rules for storing and managing data on a disk. It is like a library's classification system. Even on the same hard disk, read/write speed, stability, and features differ completely depending on the filesystem.

2. Core Comparison of the Four Filesystems

3. ext4 — Linux's Longtime Standard

Key Features

ext4 is a filesystem proven over more than 20 years. It has almost no bugs and ships as the default on every Linux distribution.

• Journaling: Writes to a journal before writing data, so recovery after an abnormal shutdown is fast
• Extent-based allocation: Groups contiguous blocks into a single extent, improving small-file handling speed
• Delayed Allocation: Buffers writes and allocates them all at once, reducing file fragmentation

Practical Commands

# Format ext4
sudo mkfs.ext4 /dev/sdb1

# Mount (default options)
sudo mount /dev/sdb1 /mnt/data

# Tune mount options — read-heavy server
sudo mount -o noatime,nodiratime,data=writeback /dev/sdb1 /mnt/data

# Register permanently in fstab
echo '/dev/sdb1 /mnt/data ext4 defaults,noatime,nodiratime 0 2' | sudo tee -a /etc/fstab

# Check disk usage
df -hT /mnt/data

# Defragmentation (ext4 has no automatic defrag, so it must be manual)
sudo e4defrag /mnt/data

# Integrity check (online possible)
sudo e2fsck -f /dev/sdb1

Best Practices

• Adding the noatime option in /etc/fstab skips unnecessary access-time recording and improves I/O performance
• data=writeback mode gives the fastest write performance but slightly increases the risk of data loss on an abnormal shutdown
• data=ordered (the default) has the best balance of stability and performance

4. XFS — The Master of Large-Volume Handling

Key Features

XFS is a high-performance filesystem developed on IRIX, optimized for large databases like MySQL/PostgreSQL and for handling large media files.

• Allocation Groups (AG): Divides the disk into independent regions to improve concurrent write performance
• B+tree-based directories: Maintains directory search speed even with millions of files
• Realtime streaming: Specialized for real-time I/O such as media streaming
• Can grow but not shrink: You can enlarge a volume but not reduce it

Practical Commands

# Format XFS
sudo mkfs.xfs /dev/sdb1

# Mount
sudo mount /dev/sdb1 /mnt/data

# Mount options — for a DB server
sudo mount -o noatime,nodiratime,logbufs=8,logbsize=256k /dev/sdb1 /mnt/data

# Register in fstab
echo '/dev/sdb1 /mnt/data xfs defaults,noatime 0 2' | sudo tee -a /etc/fstab

# Check disk usage
df -hT /mnt/data
xfs_info /dev/sdb1

# Defragmentation
sudo xfs_fsr /mnt/data

# Repair tool
sudo xfs_repair /dev/sdb1

# Online repair (with it mounted)
sudo xfs_repair -L /dev/sdb1

Best Practices

• The RHEL/CentOS/Fedora family defaults to XFS — keeping it as-is is the safest
• Combining PostgreSQL's shared_buffers with XFS's noatime maximizes DB performance
• When storing large log files, increasing the log buffer with logbufs=8 improves write performance

5. Btrfs — The Wizard of Snapshots and Data Protection

Key Features

Btrfs is based on Copy-on-Write (CoW) and supports snapshots, recovery, RAID, and real-time compression at the filesystem level. It is mainly adopted by Synology NAS and the like.

• Copy-on-Write (CoW): When modifying data it writes to a new block without touching the existing one -> snapshots are created instantly
• Real-time compression: Compresses automatically when saving files (zstd, lzo, zlib) -> saves disk space
• Btrfs RAID: Supports RAID 0/1/5/6/10 at the filesystem level without mdadm
• Subvolumes: Can be managed like independent volumes within a single partition

Practical Commands

# Format Btrfs (single, no RAID)
sudo mkfs.btrfs /dev/sdb1

# Format Btrfs (RAID 1 — mirroring)
sudo mkfs.btrfs -d raid1 -m raid1 /dev/sdb1 /dev/sdc1

# Mount
sudo mount /dev/sdb1 /mnt/data

# Register in fstab
echo '/dev/sdb1 /mnt/data btrfs defaults,noatime 0 2' | sudo tee -a /etc/fstab

# Create a subvolume
sudo btrfs subvolume create /mnt/data/@home
sudo btrfs subvolume create /mnt/data/@snapshots

# Create a snapshot (finishes in 1 second)
sudo btrfs subvolume snapshot /mnt/data/@home /mnt/data/@snapshots/home-$(date +%Y%m%d)

# List snapshots
sudo btrfs subvolume list /mnt/data

# Delete a snapshot
sudo btrfs subvolume delete /mnt/data/@snapshots/home-20260923

# Check real-time compression
sudo btrfs filesystem defragment -r -czstd /mnt/data

# Check disk usage (per subvolume)
sudo btrfs filesystem usage /mnt/data

# Integrity check
sudo btrfs scrub start /mnt/data
sudo btrfs scrub status /mnt/data

Best Practices

• A snapshot is not a backup. If the hard disk physically fails, the snapshot dies with it -> always pair it with an external backup
• The noatime option, combined with CoW, greatly improves I/O performance
• zstd compression slightly increases CPU usage but saves 30-50% of disk space

6. ZFS — The Enterprise Endgame

Key Features

ZFS is a filesystem with data integrity, powerful RAID, and large-scale caching built in. It is optimal for bundling several hard disks into safe, enormous storage.

• Checksum-based integrity: Computes a checksum for every block and automatically detects data corruption
• Self-healing: In a mirror/RAID-Z configuration, automatically recovers corrupted blocks from good ones
• ARC (Adaptive Replacement Cache): Uses RAM as a disk cache to maximize repeated-read performance
• ZFS on Linux: Not built into the kernel due to licensing, so a separate install is needed

Practical Commands

# Install ZFS (Ubuntu/Debian)
sudo apt install zfsutils-linux

# Create a ZFS pool (single disk)
sudo zpool create datapool /dev/sdb1

# Create a ZFS pool (RAID-Z1 — 1-disk parity)
sudo zpool create datapool raidz1 /dev/sdb1 /dev/sdc1 /dev/sdd1

# Create a ZFS pool (RAID-Z2 — 2-disk parity)
sudo zpool create datapool raidz2 /dev/sdb1 /dev/sdc1 /dev/sdd1 /dev/sde1

# Check pool status
zpool status datapool

# Check pool usage
zpool list datapool
zfs list

# Create a snapshot
sudo zfs snapshot datapool/home@backup-20260923

# List snapshots
sudo zfs list -t snapshot

# Recover from a snapshot
sudo zfs rollback datapool/home@backup-20260923

# Automatic snapshots (zfs-auto-snapshot)
sudo apt install zfs-auto-snapshot
# Automatic hourly snapshots (keep up to 24)
sudo zfs set com.sun:auto-snapshot:hourly=true datapool/home

# ZFS pool online repair
sudo zpool scrub datapool

# Add a disk (online expansion)
sudo zpool add datapool /dev/sdf1

Best Practices

• ZFS consumes a lot of RAM. At least 8GB, preferably 16GB or more, is recommended
• Monitoring the ARC cache hit rate with the arcstat command lets you identify disk I/O bottlenecks
• RAID-Z1 is only meaningful with 3 or more disks. With 2 disks, mirroring is safer

7. Performance Benchmark Comparison

A general performance comparison in the same NVMe SSD environment. Actual performance varies with hardware, workload, and options.

Key point: Btrfs's CoW creates only a pointer when copying a file, with no physical copy, so even a multi-GB file is copied instantly. zstd compression adds a little CPU load but saves 30-50% of disk space.

8. Practical Selection Guide

Case 1: "I don't know, I just want the most stable" -> ext4

For a general web server, a light application server, or a personal Linux PC, choose ext4. Thanks to its decades of proven stability, it has the lowest chance of data corruption even when the system goes down unexpectedly.

# The safest basic setup
sudo mkfs.ext4 /dev/sdb1
sudo mount -o defaults,noatime,ordered /dev/sdb1 /mnt/data

Case 2: "I deal with large databases or media files" -> XFS

For large DB servers such as MySQL/PostgreSQL, servers where log data accumulates in gigabytes, or handling video files, XFS is the answer.

# Optimal setup for a DB server
sudo mkfs.xfs /dev/sdb1
sudo mount -o noatime,logbufs=8,logbsize=256k /dev/sdb1 /var/lib/mysql

Case 3: "Frequent backups and preventing data loss matter" -> Btrfs

For environments that create and delete virtual machines often, or that must back up the filesystem hourly for ransomware preparedness, Btrfs is useful.

# Optimal setup for NAS/backup
sudo mkfs.btrfs /dev/sdb1
sudo mount -o defaults,noatime,compress=zstd /dev/sdb1 /mnt/data
sudo btrfs subvolume create /mnt/data/@data
sudo btrfs subvolume create /mnt/data/@snapshots

# Automatic snapshot script (crontab)
echo '0 * * * * root btrfs subvolume snapshot /mnt/data/@data /mnt/data/@snapshots/data-$(date +\%Y\%m\%d\%H)' > /etc/cron.d/btrfs-snapshot

Case 4: "A dozens-of-TB professional backup/storage server" -> ZFS

If you want to bundle several hard disks into safe, enormous storage, ZFS is recommended.

# Optimal setup for a storage server
sudo zpool create -o ashift=12 datapool raidz2 /dev/sd{b,c,d,e}
sudo zfs create -o compress=lz4 -o atime=off datapool/data
sudo zfs create -o compress=zstd datapool/backup
sudo zfs set com.sun:auto-snapshot:daily=true datapool/data

9. The 2026 Recommendation Formula

Final tip: A filesystem is hard to change once formatted. Before adopting one, always understand your own workload (I/O pattern, data size, backup requirements) and then choose.