Re: merge>hash>loop

On Wed, Apr 19, 2006 at 04:47:40PM +1200, Mark Kirkwood wrote:
> Jim C. Nasby wrote:
> >On Tue, Apr 18, 2006 at 06:22:26PM -0400, Tom Lane wrote:
> >>"Jim C. Nasby"  writes:
> >>>Actually, if you run with stats_block_level turned on you have a
> >>>first-order approximation of what is and isn't cached.
> >>Only if those stats decayed (pretty fast) with time; which they don't.
> >
> >Good point. :/ I'm guessing there's no easy way to see how many blocks
> >for a given relation are in shared memory, either...
> 
> contrib/pg_buffercache will tell you this - what buffers from what 
> relation are in shared_buffers (if you want to interrogate the os file 

So theoretically with that code we could make the cost estimator
functions more intelligent about actual query costs. Now, how you'd
actually see how those estimates improved...

> buffer cache, that's a different story - tho I've been toying with doing 
> a utility for  Freebsd that would do this).

Well, the problem is that I doubt anything that OS-specific would be
accepted into core. What we really need is some method that's
OS-agnostic...
-- 
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