Re: PG for DataWarehouse type Queries

Ow Mun Heng <Ow(dot)Mun(dot)Heng(at)wdc(dot)com> writes:
> Can anyone shed some light on this. I just would like to know if queries
> for raw data (not aggregregates) is expected to take a long time.
> Running times between 30 - 2 hours for large dataset pulls.

> Involves lots of joins on very large tables (min 1 millon rows each
> table, 300 columns per table)

> Joins are done in the form of Left joins (sometimes on the same tables,
> due to normalisation)

> Is 30min - 2hours too long or is this considered "normal"??

This question is nearly content-free. How many joins is "lots"? How
many result rows are you expecting? What PG version are you using?
What have you got work_mem set to? What does EXPLAIN say the plan is
(EXPLAIN ANALYZE output would be better, but you'll have to wait for the
results...)?

regards, tom lane