Re: Performance with new nested-xacts code

On Thu, Jul 01, 2004 at 12:21:55AM -0400, Tom Lane wrote:
> I was all set to launch into a diatribe about the half dozen performance
> issues I think we *must* fix in the new nested-transactions code,

I completely agree, of course.

> This brought me up short. I sure as heck do not see anything in that
> patch that would represent a performance gain over before, especially
> not in the very vanilla-flavor cases exercised by pgbench. Do you see
> an explanation? I'm a bit worried that we've managed to dike out some
> essential operation or other...

Nope, nothing; in fact, I think the only thing that changed in the
normal (no subxacts) code path are the tqual.c tests, and those ought to
be more expensive than before, not less ...

The only thing that changed somewhat is the new GUC code you wrote, but
I doubt that could account for a 10% performance increase.

> Can anyone else reproduce these results? The test case I'm using is
> pgbench -i -s 10 bench
> followed by repeated
> pgbench -c 5 -t 1000 bench

I measure the same in 7 runs:

Yesterday Today
241.37 243.02
215.3 228.58
235.13 225.57
194.72 253.23
215.98 272.14
239.15 262.35
220.17 249.42
Min 194.72 225.57
Max 241.37 272.14
Avg 223.12 247.76

--
Alvaro Herrera (<alvherre[a]dcc.uchile.cl>)
"Por suerte hoy explotó el califont porque si no me habría muerto
de aburrido" (Papelucho)