DWARFS

Revision as of 09:28, 17 August 2025

Deduplicating Warp-speed Advanced Read-only File System (DWARFS) is a read-only filesystem that optionally supports no compression (via --compress-level=0 or --compression=null for instance) and no deduplication (via --file-hash none. It is developed by Marcus Holland-Moritz and when compared to Squashfs, DWARFS also offers a choice of hashing algorithms,^[2] as well as a tool for checking DWARFS image via dwarfsck. At maximum compression levels using LZMA, DWARFS (using --compression=lzma:level=9:extreme) can produce smaller file size compared to SquashFS with the rough equivalent of using XZ as -comp xz.

See also Squashfs.

Discussion

The aim with this software project is to ultimately create a compressed, deduplicated, read-only file system. While it is possible to disable all except for read-only, disabling it defeats the purpose.

DWARFS software project also intends to compete (performance-wise) with SquashFS at file system creation, in that a given directory, folder, or path is scanned, hashed, (optionally categorized) before the contents are compressed, adding only the unique copies. SquashFS tends to add and hash files as it creates the file system, similar to how 7-Zip or ZIP files for instance are created, with the apparent ability to detect and avoid adding duplicate files during creation process that is unique to the likes of SquashFS and DWARFS, which however is considered slower, generally double-handling compared to DWARFS which does these during the scanning phase.

The developer's original motivation was with, "several hundred different versions of Perl that were taking up something around 30 gigabytes of disk space" and that there, "was unwilling to spend more than 10% of my hard drive keeping them around for when I happened to need them."^[3] This may put the project in line for competing with git, which is a Versioning Control System (VCS), that is used to record changes to a given project at a more atomic level, compressing (and using zlib) changes with each commit, however, unlike git which introduces rather steep learning curves, contrasting to just adding several versions of the same software project into an "archive" is arguably more trivial to do. Ultimately, DWARFS, like SquashFS has many other potential use-cases, rather than whatever they were intentionally designed for, making it a somewhat popular choice in the race against other forms of archivers and software/file distribution methods, without requiring the use of bespoke compression software that happens to be ranked among the top in Matt Mahoney's data compression benchmarks.

Identification

DWARFS files begins with the hexadecimal 44 57 41 52 46 53 which translates to "DWARFS" in ASCII.^[4]

Examples

Create an extremely compressed DwarFS image, without history, all root owned contents, sha3-512 hash algorithm, idle task (lowest priority) from present directory to example.dwarfs:

 mkdwarfs --input . --output=example.dwarfs --block-size-bits=26 --compression=lzma:level=9:extreme --compress-niceness=10 --schema-compression=lzma:level=9:extreme \
 --metadata-compression=lzma:level=9:extreme --no-history --pack-metadata=all,force --file-hash=sha3-512 --set-owner=0 --set-group=0 --no-history-timestamp \
 --no-create-timestamp --no-history-command-line

Choice of compression available

As of version 0.12.4, the following compressors are available:

Choice of hash for duplication detection available

Since 0.7.0-RC1 introduced alternatives,^[13] as of version 0.12.4, the following hash algorithms are available:

Links

• https://github.com/mhx/dwarfs

References

1. ↑ Release 0.1.0 - GitHub
2. ↑ Choice of hash for duplicate detection #92 - GitHub
3. ↑ History section - DWARFS - GitHub
4. ↑ 0000449: Add magic for the DWARFS compressed file system format - bugs.astron.com
5. ↑ null.cpp (line 110) - DWARFS - GitHub
6. ↑ lzma.cpp (line 413) - DWARFS - GitHub
7. ↑ zstd.cpp (line 179) - DWARFS - GitHub
8. ↑ lz4.cpp (line 186) - DWARFS - GitHub
9. ↑ brotli.cpp (line 169) - DWARFS - GitHub
10. ↑ flac.cpp (line 496) - DWARFS - GitHub
11. ↑ rice.hpp (lines 7-9) - compression-algorithms - GitHub
12. ↑ ricepp.cpp (line 260) - DWARFS - GitHub
13. ↑ Choice of hash for duplicate detection #92 post #5 - DWARFS - GitHub
14. ↑ Choice of hash for duplicate detection #92, post #5 - DWARFS - GitHub
15. ↑ Commit de5ec99 - test/checksum_test.cpp (line 55) - test(checksum): add checksum tests - DWARFS - GitHub
16. ↑ Choice of hash for duplicate detection #92, post #1 - DWARFS - GitHub
17. ↑ Commit afbd85e - fix(checksum): disable extended output algorithms (e.g. shake(128|256)) - DWARFS - GitHub
18. ↑ Commit afbd85e - fix(checksum): disable extended output algorithms (e.g. shake(128|256)) - DWARFS - GitHub
19. ↑ Commit 7ded26d - feat(checksum): add hexdigest() method - DWARFS - GitHub
20. ↑ v0.7.0-RC1 - DWARFS releases by tag - GitHub