On Thu, May 13, 2021 at 09:47:02PM -0400, Tom Lane wrote:
> Julien Rouhaud <rjuju123@gmail.com> writes:
> > On Fri, May 14, 2021 at 3:12 AM Bruce Momjian <bruce@momjian.us> wrote:
> >> I was surprised it was ~2%.
>
> > Just to be clear, the 2% was a worst case scenario, ie. a very fast
> > read-only query on small data returning a single row. As soon as you
> > get something more realistic / expensive the overhead goes away.
>
> Of course, for plenty of people that IS the realistic scenario that
> they care about max performance for.
I'm not arguing that the scenario is unrealistic. I'm arguing that retrieving
the first row of a join between pg_class and pg_attribute on an otherwise
vanilla database may not be the most representative workload, especially when
you take into account that it was done on hardware that still took 3 ms to do
that.
Unfortunately my laptop is pretty old and has already proven multiple time to
give unreliable benchmark results, so I'm not confident at all that those 2%
are even real outside of my machine.