Mike,

> There are 2 more experiments that would definitely be worth running:
>
> 1) with the current benchmark config and XFS filesystem, increase the log
> size to maximum (32768 blocks)
>
> 2) With ext3 and the 400M journal, increase the running time of the
> benchmark to say 600 secs?


I've combined the results into:

http://samba.org/~tridge/xattr_resul...xt3-tuning.png

The 3 curves are:

- the xfs-512 results with default journal size (from an earlier
graph, included for easy comparison)

- xfs-512 results with a journal of size 32768 blocks

- ext3 results with inode size 256 and journal 400M (and the patch to
allow ext3 to take advantage of large inodes).

The latter two results are with 600 second runs.

> Have theory, need confirmation!


The larger journal certainly helps xfs a lot, though not quite as much
as ext3.

Please let me know how well the results match your theory!

Cheers, Tridge