At 05:48 PM 12/29/00 -0500, der Mouse wrote:
>Are you sure of that? [MSCP set up overhead]  I've seen performance
>problems like this with reason to think they're actually due to each
>transfer just barely missing the sector it wants and hence having to
>wait for a full disk rotation.  (Of course, depending on how smart
>the disk in question is, this may not be plausible.)


Sure? No, never sure :-) But this is the evidence that suggests it:
         1) The disk in question is a Seagate Barracuda with a 1Mb
            buffer (ie it's pretty smart about such things)
         2) When using 512 byte blocks the activity light pulses at
            about 10hz when using 16K blocks it stays full on.
         3) The disk "geometry" is made up (based on number of
            sectors reported, hence little relation to the actual
            geometry.)

I did try re-formatting the drive with the "drive optimum interleave" 
(basically leave it up to the drive to do the Right Thing) and it got a 
fractionally bit better (from 94K to 107K) so there may have been an 
interleave component but again I think I've optimized it out.

--Chuck