On Fri, Jan 27, 2006 at 08:00:17PM +0100, Manuel Bouyer wrote:
> 
> Not really. Eventually you can try different strip size and see which
> work better for your usage. The block size tuning isn't dependant on
> raidframe, but rather on the type of data stored on disk (lots of small
> files vs a few large files)

That's not entirely so: you want blocksize * maxcontig to work out to
the stripe size or a multiple of it.  So, for instance, if you have a
four-disk RAID5, you want to be writing 48K at a time, which means a
16K FFS block size and maxcontig=3.

However, FFS has no good way to ensure that data is stripe aligned, so
you will still see strange artifacts.

Thor