Just to give my first experiences of PostgreSQl after working
several days with this tool:
a)
I´ve done some benchmarks about the behaviour of PostgreSQL, when working
with tables with many, many rows. I´ve created a table with two columns
(varchar and integer), both with indexes. The insert speed was about
533 insert/seconds at the beginning, but slows down to 15 insert/seconds
when having 600000 rows. It´s not a linear movement, but a degressive
slow down (first fast, later slower).
PostgreSQL has the possibility of three hash methods. Which one is
suited for which cases ? Just to get an impression ?
b)
I´ve noticed, that "vacuumdb" can help, but I noticed, that it seems
to be better not to run "vacuumdb" while another program is doing heavy
work in some tables. I´ve seen several deadlocks between my test
programs and "vacuumdb".
c)
The size of the index files grows and grows and grows. "vacuumdb" does
not reduce the size of the index files, though in "verbose" mode it
seems to tell me, that it deleted several entries in the index files.
New index changes seems to be appended to the index files :-(
The only way to get rid of this seems to drop all indexes and create
them again ... improving the speed also.
This is a critical point ...
d)
The library "libpg" is very nice ! Small and it does it´s work !
What I would like to know. has anybody working databases using PostgreSQL
with sizes more than 500 MByte and is happy with it ??? We consider it
using for research projects starting in autumn, but we would like to
know, how it behaves with large databases ?
Marten