On Mon, Sep 9, 2013 at 07:39:00PM -0400, Bruce Momjian wrote:
> > In the case of tablespaces, I should have thought you could keep a
> > hash table of the names and just store an entry id in the table
> > structure. But that's just my speculation without actually looking
> > at the code, so don't take my word for it :-)
>
> Yes, please feel free to improve the code. I improved pg_upgrade CPU
> usage for a lerge number of objects, but never thought to look at memory
> usage. It would be a big win to just palloc/pfree the memory, rather
> than allocate tones of memory. If you don't get to it, I will in a few
> weeks.
Thanks you for pointing out this problem. I have researched the cause
and the major problem was that I was allocating the maximum path length
in a struct rather than allocating just the length I needed, and was not
reusing string pointers that I knew were not going to change.
The updated attached patch significantly decreases memory consumption:
tables orig patch % decrease
----
1 27,168 kB 27,168 kB 0
1k 46,136 kB 27,920 kB 39
2k 65,224 kB 28,796 kB 56
4k 103,276 kB 30,472 kB 70
8k 179,512 kB 33,900 kB 81
16k 331,860 kB 40,788 kB 88
32k 636,544 kB 54,572 kB 91
64k 1,245,920 kB 81,876 kB 93
As you can see, a database with 64k tables shows a 93% decrease in
memory use. I plan to apply this for PG 9.4.
--
Bruce Momjian <bruce@momjian.us> http://momjian.us
EnterpriseDB http://enterprisedb.com
+ Everyone has their own god. +
