Re: Parallel INSERT (INTO ... SELECT ...)

On Mon, Jan 18, 2021 at 1:02 PM Tang, Haiying
<tanghy(dot)fnst(at)cn(dot)fujitsu(dot)com> wrote:
>
> > From: Amit Kapila <amit(dot)kapila16(at)gmail(dot)com>
> > > Can we test cases when we have few rows in the Select table (say
> > > 1000) and there 500 or 1000 partitions. In that case, we won't
> > > select parallelism but we have to pay the price of checking
> > > parallel-safety of all partitions. Can you check this with 100, 200,
> > > 500, 1000 partitions table?
> >
> > I also wanted to see such an extreme(?) case. The 1,000 rows is not
> > the count per partition but the total count of all partitions.e.g.,
> > when # of partitions is 100, # of rows per partition is 10.
>
> Below results are in serial plan which select table total rows are 1,000. The Excution Time + Planning Time is still less than unpatched.
> (does this patch make some optimizes in serial insert? I'm a little confused here, Because the patched execution time is less than unpatched, but I didn't find information in commit messages about it. If I missed something, please kindly let me know.)
>

I don't think the patch should have any impact on the serial case. I
think you can try to repeat each test 3 times both with and without a
patch and take the median of the three.

--
With Regards,
Amit Kapila.