Saturday, October 24, 2009

Goodbye ZFS, Hello something else...

Apple has shut down the source repository for the Mac OS X port of ZFS, leading to a flurry of reports that the project has been canned, apparently for legal rather than technical reasons.

While this may leave Apple without an advanced project to develop a next-generation operating system, it certainly doesn't leave them without a wide array of examples nor without a choice of starting points.

Apple could, of course, start from scratch, designing a file system to suit their particular circumstances, perhaps even one that would scale from the smallest portable devices to the largest Xsan installations, with plenty of elbow room for future increases in bit density.

They could also start with existing projects, such as UFS2, associated with the FreeBSD project, or maybe HAMMER, associated with DragonFly BSD, an offshoot of FreeBSD.

Many of us were looking forward to ZFS, not because it was ZFS in particular, but because it would presumably have been both faster and more stable than HFS+ (starting about halfway down). Frankly, the speed advantage of one file system design over another mostly goes away as you move from hard disks to solid state memory, since there's no time-consuming repositioning of read/write heads in solid state memory. (There can still be minor differences in speed due to code efficiency, but these pale by comparison.) While most desktop and laptop computers continue to come with hard disks, the cost/capacity of solid state memory is coming down rapidly, and it has already displaced hard disks in the most portable devices. Speed is on the way to becoming a non-issue.

What's mainly left is security (control over who has access to what), stability (partly a matter of redundancy), a logical structure that contributes to rather than interferes with keeping content organized and making it accessible, and the completeness of metadata (file name, file type, file owner, creator, creating program, date of creation, source URI, ...).

Apple might also wish to integrate (aspects of) its Spotlight and/or Core Data technologies into the file system. I'm not enough of an adept to know whether there's anything to be gained in this, but it seems possible.

All in all, it seems like a win that Apple is looking to its own resources for a next-generation replacement for HFS+.

No comments: