Possible evidence of performance regression for 8.1-S (vs. 7.1)

[Ivan Voras]

On 10/27/10 12:55, David Wolfskill wrote:

> That *is* a problem, as I cannot justify a migration to a branch
> of FreeBSD that imposes about a 23% penalty in elapsed time on this
> workload.  I want folks at work to have more reason to want to use
> (newer branches of) FreeBSD, not less.

That is understandable, I'm only saying that given the symptoms - some
"waiting" that is not accounted in system or user time and with NFS
mostly ruled out (maybe a better test would be to extract a local
tarball *to* the NFS server), the next suspect is the disk system. In
case you never tried "diskinfo -vt", its results look like this:

	outside:       102400 kbytes in   1.222998 sec =    83729 kbytes/sec
	middle:        102400 kbytes in   1.448580 sec =    70690 kbytes/sec
	inside:        102400 kbytes in   2.531694 sec =    40447 kbytes/sec

note 2x drop in performance between outer and inner tracks. It's not
certain, especially since your results are repeatable and you always use
exactly the same file system, but it's possible. Have you tried the
sysctls I've posted?