alexandriafs

The AleXandria File System

The AleXandriaFS is designed for modern storage devices without any moving parts (SSDs, SD cards, USB sticks, etc.), provides all the features you'd expect from an enterprise-grade file system (write cycle protection, access control lists, extended attributes, journaling, encryption etc.) and it is very efficient, flexible, and very performant. Furthermore AleXandriaFS also made it its goal to use as small portion of the storage as possible for its meta data.

The low-level on-disk format can be found in the AleXandriaFS Specification.

The linux directory contains a Linux kernel module providing mount -t axfs. More information can be found here on how to use it. (WIP, read-only for now)

The src directory on the other hand contains dependency-free and stand alone user space tools like mkfs.axfs, fsck.axfs etc., these are implemented both for the POSIX and for the WIN32 API. These work on device and image files as well. The tool.axfs is a no root privilege required swiss army knife tool (much like mtools), but you can also use the fuse.axfs FUSE driver.

The entire file system is implemented in two ANSI C header files: axfs.h contains all the on-disk structures, and libaxfs.h is an stb-style single header library with all the algorithms and a high level user API.

Limitations

Not really a good phrase, because using this file system feels like if it were truly unlimited.

Property	Biggest allowed
logical block size	2^264 (but that's not practical. The default is 4k, 64k being a more likely limit)
one volume size	2^64 blocks (with 4k blocks it's 2^76 bytes, 64 Zetta, or ten thousands of million Terrabytes)
one file	2^80 bytes (also depends on block size and the actual volume capacity)
longest file name	246 bytes (less in characters, because some UTF-8 characters are multi-byte)
O(log2) access directories	2^32 bytes (ca. 16 million entries, if all file names are 246 bytes, more if they are shorter)
entries in a directory	2^32 (less than 4 billion, because unique names force file names to grow)
blocks in one extent	2^64 - 1
number of attributes	2^32 in total, one compressed block (with 4k blocks about 500 attributes) per inode

Most important aspects are all handled in a flexible way, for example there's absolutely no limit on number of inodes, and you can resize the volume or just the journal buffer on-the-fly. The biggest advantage of AXFS is scalability, that the size of the required meta data does not grow as the volume size increases like with most other file systems.

License

AleXandriaFS as well as its supporting library is Free and Open Source Software, licensed under the terms of GPL version 3 or (at your opinion) any later version. See the LICENSE file for details.