parimala escreveu:
[Don't repeat your answer. It's a PITA to receive multiple identical copies]
> We are using Postgres 8.3.7 in our java application. We are doing
> performances tuning and load testing in our setup. we have noticed that
> ,some of our queries to the database taking long time to return the
> results.Please find our setup details belows.
>
> We observed that postgres is running in windows is slower than the linux .
>
That is true and it will be for quite some time. Windows port is very recent
if we compare it with the long road Unix support.
> Postgresql configuration:
>
> shared_buffers: 1 GB
I don't use Windows but I read some Windows users saying that it isn't
appropriate to set the shared_buffers too high. Take a look at the archives.
> Effective_cache_size: 2GB
> fsync: off (even we tested this parameter is on ,we observed the same
> slowness )
>
What about the other parameters that are different from default (uncommented
parameters)? Also, don't turn off the fsync unless you're pretty sure about
the consequences.
> We have 300k row's in PolledData Table.In each STATSDATA table ,we have
> almost 12 to 13 million rows. Every one minute interval ,we insert data
> into to STATSDATA table. In our application ,we use insert and select
> query to STATSDATA table at regular interval. Please let us know why the
> below query takes more time to return the results. is there any thing we
> need to do to tune the postgres database ?
>
It seems very strange that your queries are not using the indexes. Do you have
autovacuum turn on? Do you recently analyze your tables?
> Merge Cond: (statsdata8_21_2009.pollid =
> ((polleddata.id)::numeric))
Out of curiosity, why does foreign key have different datatype of its primary key?
--
Euler Taveira de Oliveira
http://www.timbira.com/
