Bruce Momjian wrote:
>
> >
> > Oracle:
> > I needed 24 seconds and 1937 microseconds for the insert test.
> > I needed 1 seconds and 600641 microseconds for the selection&projection test.
> > I needed 1 seconds and 703673 microseconds for the join test.
> > I needed 2 seconds and 404709 microseconds for the update test.
> >
> > PostgreSQL with -F:
> > I needed 10 seconds and 297716 microseconds for the insert test.
> > I needed 28 seconds and 964208 microseconds for the selection&projection test.
> > I needed 83 seconds and 931762 microseconds for the join test.
> > I needed 0 seconds and 588390 microseconds for the update test.
>
> This seems strange. Is a vacuum being done so the optimizer knows how
> large each table is?
You're right, Bruce. I don't see VACUUM in Michael' programm.
This also affects the second test (selection&projection): during
select backend updates data pages to set commit/abort statuses
for tuples' xactions.
Vadim