Choosing a Filesystem

A filesystem determines how data is stored, organized, and retrieved on your disk. Choosing the right one affects performance, reliability, and the features available to you. On Linux, you have three main choices: ext4, XFS, and Btrfs. Each excels in different scenarios.

1. Quick Comparison

2. ext4 — The Reliable Default

ext4 (Fourth Extended Filesystem) is the most widely used Linux filesystem. It evolved from ext3 and has been the default on Debian and Ubuntu for over 15 years. It's the safest, most predictable choice for general-purpose workloads.

Strengths

• Battle-tested reliability — used on billions of systems, extremely stable
• Excellent small-file performance — great for web serving, mail, and mixed workloads
• Low overhead — minimal CPU and RAM usage
• Offline shrink support — the only one of the three that can be shrunk
• Widest tooling support — every recovery and repair tool works with ext4

Limitations

• No built-in snapshots or compression
• Doesn't checksum file data (only metadata)
• Slower than XFS for very large sequential writes and highly parallel I/O

3. XFS — The Performance Powerhouse

XFS was originally developed by SGI in 1993 for their IRIX workstations handling massive video and scientific data. It's designed from the ground up for high throughput, large files, and parallel I/O. RHEL, AlmaLinux, and Rocky Linux use it as their default.

Strengths

• Excellent large-file performance — streaming, video, backups, databases with large tablespaces
• Best parallel I/O — scales linearly with CPU cores for concurrent operations
• Delayed allocation — optimizes write placement, reducing fragmentation
• Reflink copies — instant copy-on-write file clones (since Linux 4.9+)
• Online defragmentation — xfs_fsr while mounted

Limitations

• Cannot be shrunk — only grown. Plan your initial size carefully.
• No built-in snapshots or compression
• Slightly slower than ext4 for many small files and metadata-heavy operations

4. Btrfs — The Feature-Rich Modern Option

Btrfs (B-tree Filesystem, pronounced "butter-FS") is a copy-on-write filesystem that brings features like snapshots, compression, and checksumming that traditionally required separate tools or volume managers. It's the default on Fedora and openSUSE.

Strengths

• Native snapshots — instant, space-efficient snapshots for backups and rollbacks
• Transparent compression — zstd, lzo, or zlib; saves disk space automatically
• Data checksumming — detects silent data corruption (bit rot)
• Subvolumes — logical divisions within a filesystem, each mountable independently
• Online shrink and grow — resize in both directions while mounted
• Send/receive — incrementally replicate snapshots between servers

Limitations

• RAID 5/6 is unstable — do not use Btrfs RAID 5/6 in production (RAID 1/10 are fine)
• Higher overhead — copy-on-write and checksumming use more CPU and RAM
• Fragmentation — copy-on-write can fragment heavily under random-write workloads (databases)
• Space accounting — free space reporting can be confusing with snapshots and compression

5. When to Use What

6. Formatting & Mounting

These examples use /dev/vdb as the target device — a common name for block storage volumes. Replace with your actual device.

ext4

# Format as ext4
sudo mkfs.ext4 /dev/vdb

# With a label
sudo mkfs.ext4 -L mydata /dev/vdb

# Mount
sudo mkdir -p /mnt/data
sudo mount /dev/vdb /mnt/data

# Add to /etc/fstab for persistent mount
echo '/dev/vdb /mnt/data ext4 defaults 0 2' | sudo tee -a /etc/fstab

XFS

# Format as XFS
sudo mkfs.xfs /dev/vdb

# With a label
sudo mkfs.xfs -L mydata /dev/vdb

# Mount
sudo mkdir -p /mnt/data
sudo mount /dev/vdb /mnt/data

# Add to /etc/fstab
echo '/dev/vdb /mnt/data xfs defaults 0 0' | sudo tee -a /etc/fstab

Btrfs

# Install Btrfs tools if needed
sudo apt install btrfs-progs        # Debian/Ubuntu
sudo dnf install btrfs-progs        # AlmaLinux/RHEL

# Format as Btrfs
sudo mkfs.btrfs /dev/vdb

# With a label
sudo mkfs.btrfs -L mydata /dev/vdb

# Mount
sudo mkdir -p /mnt/data
sudo mount /dev/vdb /mnt/data

# Mount with compression enabled
sudo mount -o compress=zstd /dev/vdb /mnt/data

# Add to /etc/fstab with compression
echo '/dev/vdb /mnt/data btrfs defaults,compress=zstd 0 0' | sudo tee -a /etc/fstab

Btrfs Subvolumes

# Create subvolumes
sudo btrfs subvolume create /mnt/data/@home
sudo btrfs subvolume create /mnt/data/@logs

# Mount a specific subvolume
sudo mount -o subvol=@home /dev/vdb /home
sudo mount -o subvol=@logs /dev/vdb /var/log

# List subvolumes
sudo btrfs subvolume list /mnt/data

# Create a snapshot
sudo btrfs subvolume snapshot /mnt/data/@home /mnt/data/@home-snapshot

# Create a read-only snapshot (for backups)
sudo btrfs subvolume snapshot -r /mnt/data/@home /mnt/data/@home-backup

# Show filesystem type for all mounts
df -Th

# Show filesystem type for a specific device
lsblk -f

7. Checking & Repairing

ext4

# Check (must be unmounted)
sudo umount /dev/vdb
sudo e2fsck -f /dev/vdb

# Auto-fix errors
sudo e2fsck -fy /dev/vdb

# Check and report without fixing
sudo e2fsck -n /dev/vdb

XFS

# Check (must be unmounted)
sudo umount /dev/vdb
sudo xfs_repair /dev/vdb

# Dry run (check only)
sudo xfs_repair -n /dev/vdb

Btrfs

# Online scrub (can run while mounted — checks data integrity)
sudo btrfs scrub start /mnt/data

# Check scrub status
sudo btrfs scrub status /mnt/data

# Offline check (must be unmounted)
sudo umount /dev/vdb
sudo btrfs check /dev/vdb

# Show filesystem usage
sudo btrfs filesystem usage /mnt/data

8. Performance Tuning

ext4 Tuning

# /etc/fstab — optimized ext4
/dev/vdb /mnt/data ext4 defaults,noatime,commit=60 0 2

# noatime  — don't update access time on reads (big improvement)
# commit=60 — flush data every 60s instead of 5s (less disk I/O)

XFS Tuning

# /etc/fstab — optimized XFS
/dev/vdb /mnt/data xfs defaults,noatime,logbufs=8,logbsize=256k 0 0

# noatime    — skip access time updates
# logbufs=8  — more in-memory log buffers
# logbsize=256k — larger log buffer size

Btrfs Tuning

# /etc/fstab — optimized Btrfs
/dev/vdb /mnt/data btrfs defaults,noatime,compress=zstd:3,space_cache=v2 0 0

# compress=zstd:3 — zstd compression at level 3 (good balance)
# space_cache=v2 — improved free space tracking
# noatime        — skip access time updates

# Disable COW for database directories (run BEFORE writing data)
chattr +C /mnt/data/mysql

I/O Scheduler

# Check current scheduler
cat /sys/block/vda/queue/scheduler

# For NVMe/SSD (most cloud VPS)
echo "none" | sudo tee /sys/block/vda/queue/scheduler

# For HDD
echo "mq-deadline" | sudo tee /sys/block/vda/queue/scheduler

9. Distribution Defaults

Cloud images on RamNode use the filesystem default for each distribution:

Next Steps

Now that you understand the trade-offs, here are related topics to explore: