On Wed, Jun 07, 2006 at 05:05:23PM -0400, Tom Lane wrote:
> The other thing that was bothering me was whether disk I/O might be
> undercounted by an interrupt-driven method. I kinda doubt that any
> kernel will save up N interrupts that occur while the process is blocked
> on a slow read() ... you'll probably get only one. So the whole idea
> may be unworkable.
True, but if you get to the point where you're waiting on I/O, I would
think you could afford a gettimeofday() call. Even if the block we need
is in the OS cache, it's possible that the overhead of getting it from
there is high enough that the gettimeofday() call won't matter. FWIW, it
looks like running a much larger setting for shared_buffers (like, 50%
of memory) is a good way to boost performance, and that configuration
would make it less likely that you'd do a gettimeofday just to pull data
out of the OS cache.
--
Jim C. Nasby, Sr. Engineering Consultant jnasby@pervasive.com
Pervasive Software http://pervasive.com work: 512-231-6117
vcard: http://jim.nasby.net/pervasive.vcf cell: 512-569-9461