On Wed, Apr 11, 2012 at 7:11 PM, Eyal Wilde <eyal@impactsoft.co.il> wrote:
> in a single execution postgresql may be less the twice slower than ms-sql,
> but in 20 simultaneous clients, it's about 6 times worse. why is that?
>
> the hardware is one 4-core xeon. 8GB of ram. the database size is just a few
> GB's. centos-6.2.
>
> do you think the fact that postgresql use a process per connection (instead
> of multi-threading) is inherently a weakness of postgrsql, regarding
> scale-up?
> would it be better to limit the number of connections to something like 4,
> so that executions don't interrupt each other?
What about posting some details on the tables, the 20 queries, the temp table?
I'm thinking creating so many temp tables may be hurting pgsql more
than mssql. You might want to try unlogged temp tables, which more
closely resemble mssql temp tables.