Thread: Parallel Postgresql

Parallel Postgresql

From
Walter Landry
Date:
Hello,

I have a user with a postgresql database running on a Linux cluster.
They would like to use the cluster to speed up an individual query.
I've found the product Clusgres [1], but that is more oriented towards
making multiple queries run in parallel.  I would need a single query to
be sped up.

A little searching turns up some info that Oracle, DB2, Informix, and
maybe others can do this, but I couldn't find anything for Postgresql.
If it isn't already in Postgresql, how hard would it be to implement?

Thanks,
Walter Landry
landry@osc.edu

[1] http://www.linuxlabs.com/clusgres.html


Re: Parallel Postgresql

From
"Scott Marlowe"
Date:
On Thu, 2004-07-15 at 12:08, Walter Landry wrote:
> Hello,
>
> I have a user with a postgresql database running on a Linux cluster.
> They would like to use the cluster to speed up an individual query.
> I've found the product Clusgres [1], but that is more oriented towards
> making multiple queries run in parallel.  I would need a single query to
> be sped up.
>
> A little searching turns up some info that Oracle, DB2, Informix, and
> maybe others can do this, but I couldn't find anything for Postgresql.
> If it isn't already in Postgresql, how hard would it be to implement?

The only answer to speed up individual queries in PostgreSQL right now
is to just throw a faster machine at it.  While at some future point
someone may implement the ability for pgsql to run a query in multiple
threads on different processors etc. it hasn't been done yet, and likely
won't be done soon.  PostgreSQL's main focus is on handling large
numbers of users fairly quickly, so this corner case is unlikely to get
a lot of attention.

Now, if someone were to put in the hours it would take to make it
possible for the backend to spawn extra processes to do this, I'm sure
someone might take a look at it.  But it's a pretty small gain for most
users for a lot of changes that have to be made to the way the backend
works.