Chunkfs is a new file system architecture that divides file systems up into chunks that are invidually and incrementally checked and repaired, with only occasional references to metadata in other chunks. Chunkfs is one example of repair-driven file system design, the idea that file systems should be designed from the ground up to be not only fast and efficient during "normal" operation, but also during file system check and repair.

Chunkfs was prototyped and measured several times. The conclusion is that the idea works, you can easily reduce file system repair time by a couple of orders of magnitude, and you need to design it into your system from the start. I don't plan any further work on chunkfs given that btrfs was designed with check and repair in mind from the beginning and is clearly going to be the next major Linux file system.

Useful documents and code: