Re: max_wal_senders must die

Robert Haas wrote:
> On Wed, Oct 20, 2010 at 3:40 PM, Greg Stark <gsstark(at)mit(dot)edu> wrote:
> > On Wed, Oct 20, 2010 at 6:29 AM, Robert Haas <robertmhaas(at)gmail(dot)com> wrote:
> >> Exactly. ?It doesn't take many 3-7% slowdowns to add up to being 50%
> >> or 100% slower, and that sucks. ?In fact, I'm still not convinced that
> >> we were wise to boost default_statistics_target as much as we did. ?I
> >> argued for a smaller boost at the time.
> >
> > Well we don't want to let ourselves be paralyzed by FUD so it was
> > important to identify specific concerns and then tackle those
> > concerns. Once we identified the worst-case planning cases we profiled
> > them and found that the inflection point of the curve was fairly
> > clearly above 100 but that there were cases where values below 1,000
> > caused problems. So I'm pretty happy with the evidence-based approach.
>
> The inflection point of the curve was certainly a good thing for us to
> look at but the fact remains that we took a hit on a trivial
> benchmark, and we can't afford to take too many of those.

Agreed. If people start wondering if our new major releases are perhaps
_slower_ than previous ones, we have lost a huge amount of momentum.

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

+ It's impossible for everything to be true. +