casfs

Casfs, a local content-addressable file system.

// make it work, make is fast, make it clean

Motivation

Casfs in a file system, with a very simple design. File metadata (inode table) is stored in a sqlite database (see dedicated sqlfs project), and file contents relies on a content-addressable storage. Casfs is mainly designed to be a test platform / support backend for cloudfs.

API status

Stable !

Roadmap

• Writable Inodes POC (rename, delete, mkdir)
• Initial test flow
• Proper deployment flow
• Writable/editable files
• a bit better test suite (win/linux)
• switch to dedicated project / slice cloudfs
• Readable big files
• Writable big files (continuous mode)
• Publish read-only mode

Background daemon & pending tasks

• With full test suite (e.g. winfsp/secfs test suite)
• [ ] Append file/big files (postponed)
• [ ] Embbed configuration/web browse server (rejected to cloudfs)
• [ ] Garbage collection (rejected to cloudfs ?)

Features

• Simple by design
• Unlimited file size (casfs is mostly designed to store and manage 100k files of 8GB+ - aka HD BR rips)
• Available on all platforms (linux & Windows)
• Fast (sqlite is actually fastest than most file system)
• large subset of POSIX including reading/writing files, directories, rename, symlinks, mode, uid/gid, and extended attributes
• renames do not invole any kind of server side copy
• native file deduplication - through CAS
• Compatible with existing CAS

Additional features

• nice configuration GUI
• Directroy tree snapshot / rollback / sealing (pure SQL)
• Instant file deletion (pure SQL)
• Server side TAR creation (so content duplication) - through static large object.

Related

• cloudfs main application
• s3ql python based, non CAS (but fixed block)

Credits

• 131 - author