I tend to use XFS for most Linux partitions, and Reiser4 for /tmp and filesystem databases with tens of thousands of files, as it won’t run out of inodes.
In these benchmark tests XFS performs as I would expect, but the Reiser4 and JFS results surprised me. Reiser4’s handling of actions on big trees doesn’t live up to it’s reputation, and I’ve always viewed JFS as well-built but slow, no-where near as fast as the benchmarking indicates.
I suppose this shows the value of research over conventional wisdom.
