What is Bcachefs and how does its architecture differ from Btrfs and ZFS?

Bcachefs is a modern, transactional copy-on-write (CoW) filesystem in the Linux kernel (merged in Linux 6.7). Unlike ZFS, which uses a proprietary OpenZFS module and complex SLAB memory managers, Bcachefs is native to Linux. Unlike Btrfs, which uses separate subvolume and extent B-trees that can suffer from free-space fragmentation and write holes, Bcachefs uses a unified 64-bit B-tree key-value architecture where all metadata—inodes, dirents, extents, and allocation counters—is stored as structured bkeys within a compact log-structured extent tree.

How does multi-device tiering and target grouping work in Bcachefs?

Bcachefs natively separates storage devices into target groups (e.g., target=nvme for fast SSDs and target=hdd for high-capacity drives). Operators configure foreground writes (data_replicas, foreground_target) to land on high-speed NVMe flash, while a background rebalance worker (background_target) asynchronously migrates cold, unaccessed extents down to mechanical HDDs or QLC flash without unmounting or pausing I/O.

How does Bcachefs handle extent splitting and Copy-on-Write (CoW) fragmentation?

When an existing extent is partially overwritten, Bcachefs does not overwrite the disk blocks in place. Instead, it allocates fresh buckets on the target device, writes the new payload, and splits the existing extent in the B-tree into up to three extents: the unmodified head, the newly written chunk, and the unmodified tail. Old blocks remain intact until transactions commit, providing atomic crash consistency and zero-cost snapshots.

What is the function of Bcachefs buckets and the bucket allocator?

In Bcachefs, disks are partitioned into fixed-size contiguous segments called buckets (typically 256 KB to 2 MB). Buckets serve as the fundamental unit of wear-leveling, garbage collection, and journal accounting. The bucket allocator tracks generation numbers per bucket to prevent ABA stale reference bugs and ensures copy-on-write allocation happens sequentially across erase blocks.

Linux Bcachefs Multi-Device Tiering & CoW Extent Tree Studio

Verified Specification | Standardized Formulas | Instant Precision

Secure & Private (Zero Data Retention) Free Access • No Sign-Up

Linux 6.7+ Kernel Mainline B-Tree Extents Multi-Tier SSD/HDD

Linux Bcachefs Multi-Device Tiering & CoW Studio

Simulate Bcachefs extent tree allocations, Copy-on-Write overwrite splitting, NVMe foreground to HDD background migration, and checksum scrubbing.

Foreground Target Tier

Background Migration Target

Data Replicas & Compression

Active Inode Extents

3 Extents

Total Allocated: 64 KB

Tier 1 NVMe Cache

48 KB (75%)

Wear Buckets: 3 / 16

Tier 2 Cold HDD Pool

16 KB (25%)

Rebalance status: Idle

Virtual Inode Logical Address Space (Offset 0x0000 → EOF)

Click any extent block to inspect bkey descriptor & pointers

Tier 1 (NVMe Foreground)

Tier 2 (HDD Demoted)

CoW Overwritten Target

bkey Extent Descriptor

Select an extent block above to display its bkey pos: (inode:offset), size, CRC32C checksum, and physical disk pointers.

B-Tree Inode Extents Log (BTREE_ID_extents)

Every extent is a self-describing 64-bit bkey in the unified B-tree index.

Journal & Transaction Log

Journal Sequence (seq): 0x00000042

Transaction Status: COMMITTED

Pending CoW Splits: 0 pending extents

Scrub Status: Checksums Verified (0 errors)

Atomic journal entries are written to circular ring-buffers on flash before btree updates are finalized.

# Format multi-tier Bcachefs filesystem with NVMe foreground & HDD background
bcachefs format   --label=nvme.nvme0n1 /dev/nvme0n1   --label=nvme.nvme1n1 /dev/nvme1n1   --label=hdd.sda /dev/sda   --label=hdd.sdb /dev/sdb   --foreground_target=nvme   --promote_target=nvme   --background_target=hdd   --data_replicas=2   --metadata_replicas=2   --compression=zstd

# Mount with active tiering policy
mount -t bcachefs /dev/nvme0n1:/dev/nvme1n1:/dev/sda:/dev/sdb /mnt/bcachefs   -o foreground_target=nvme,background_target=hdd

# Trigger manual background rebalance migration for cold blocks
echo 1 > /sys/fs/bcachefs/<uuid>/internal/trigger_rebalance

# Inspect data distribution across tiers
bcachefs data show /mnt/bcachefs

# Dynamically set compression per directory
bcachefs subvolume create /mnt/bcachefs/archive
bcachefs setattr --compression=zstd:15 --background_target=hdd /mnt/bcachefs/archive

# Online cryptographic checksum verification and scrubbing
bcachefs data scrub /mnt/bcachefs

# Offline deep filesystem consistency check
bcachefs fsck -v /dev/nvme0n1:/dev/nvme1n1:/dev/sda:/dev/sdb

# Inspect B-tree extents and allocation buckets
bcachefs show-super -l /dev/nvme0n1

Linux Bcachefs Multi-Device Tiering & CoW Studio

Virtual Inode Logical Address Space (Offset 0x0000 → EOF)

B-Tree Inode Extents Log (BTREE_ID_extents)

Journal & Transaction Log

🔍 Companion Workbenches & Verification Tools

Linux Continuous Profiling, Stack Unwinding & FlameGraph Studio

Redis Architecture, Memory Optimization & Eviction Policy Studio

Linux io_uring Fixed Buffers & Zero-Copy Network TX/RX Studio

Database WAL, ARIES Recovery Algorithm & Checkpointing Studio

Why Smart Users Are Switching to This Free Toolkit

The Browser Feature Most People Never Use

Stop Paying for Software You Can Get Free

File Conversion Tricks That Save Hours Every Week

This Free Tool Replaced 5 Paid Apps for Our Team

What Your Browser Can Do Without Any Extensions