Matthias Buelow <mkb@mukappabeta.de>  wrote:
 > is there something foul with RAIDframe, concerning system CPU time?
 > I have a directory with ~1000 entries on a RAID5, and a ls -l takes
 > 30 seconds wall clock, of which 17 (!) seconds are system time, on
 > an otherwise only lightly loaded machine.  On a single disk of the
 > same type, the same thing completes in a tenth of a second.
 > Also, load average is always at least at 0.80+, even if the machine
 > is doing very little, and jumps up over 1.5 as soon as the RAIDs
 > get accessed (top etc. show almost no processes running during the
 > interval and display 80-100% time idle, so there seems to be something
 > odd with the load calculation aswell).


I saw a similar thing. Turns out the default geometry that raidframe
supplies for the raid-set affects performance drastically.  I saw
100% cpu utilization and taking 30 seconds to do a 'mkdir'.

Bill Sommerfeld checked in a fix for this after 1.5.

Check the geometry of the 'disk' in the raid disklabel.  Set it 
to something sane.