Re: Further pg_upgrade analysis for many tables

On Tue, Nov 27, 2012 at 09:35:10PM -0800, Jeff Janes wrote:
> > I tested custom format with pg_restore -j and -1, as well as text
> > restore. The winner was pg_dump -Fc | pg_restore -1;
>
> I don't have the numbers at hand, but if my relcache patch is
> accepted, then "-1" stops being faster.
>
> -1 gets rid of the AtOEXAct relcache N^2 behavior, but at the cost of
> invoking a different N^2, that one in the stats system.

OK, here are the testing results:

#tbls git -1 AtOEXAct both
1 11.06 13.06 10.99 13.20
1000 21.71 22.92 22.20 22.51
2000 32.86 31.09 32.51 31.62
4000 55.22 49.96 52.50 49.99
8000 105.34 82.10 95.32 82.94
16000 223.67 164.27 187.40 159.53
32000 543.93 324.63 366.44 317.93
64000 1697.14 791.82 767.32 752.57

Up to 2k, they are all similar. 4k & 8k have the -1 patch as a win, and
16k+ really need both patches.

I will continue working on the -1 patch, and hopefully we can get your
AtOEXAct patch in soon. Is someone reviewing that?

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

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