Re: Auto-tuning work_mem and maintenance_work_mem

On Wed, Oct 9, 2013 at 02:34:19PM -0400, Robert Haas wrote:
> On Wed, Oct 9, 2013 at 2:28 PM, Bruce Momjian <bruce(at)momjian(dot)us> wrote:
> > On Wed, Oct 9, 2013 at 01:49:23PM -0400, Robert Haas wrote:
> >> > Having really bad defaults so everyone knows they are bad really isn't
> >> > user-friendly because the only people who know they are really bad are
> >> > the people who are tuning them already. Again, we need to think of the
> >> > typical user, not us.
> >>
> >> I think a typical user will be happier if we simply raise the default
> >> rather than stick in an auto-tuning formula that's largely wishful
> >> thinking. You're welcome to disagree, but you neither quoted nor
> >> responded to my points about the sorts of scenarios in which that
> >> might cause surprising and hard-to-debug results.
> >
> > Well, pointing out that is will be negative for some users (which I
> > agree) doesn't refute that it will be better for most users.
>
> That is, of course, true. But I don't think you've made any argument
> that the pros exceed the cons, or that the formula will in general be
> accurate. It's massive simpler than what Josh says he uses, for
> example, and he's not making the completely silly assumption that
> available RAM is 4 * shared_buffers. An auto-tuning formula that's
> completely inaccurate probably won't be better for most users.

I disagree. I think we can get a forumla that is certainly better than
a fixed value. I think the examples I have shown do have better value
than a default fixed value. I am open to whatever forumula people think
is best, but I can't see how a fixed value is a win in general.

--
Bruce Momjian <bruce(at)momjian(dot)us> http://momjian.us
EnterpriseDB http://enterprisedb.com

+ Everyone has their own god. +