Re: Parallel execution strategy

From: <jerome.bellet_at_altran.com>
Date: Tue, 18 Mar 2008 01:59:02 -0700 (PDT)
Message-ID: <a7ad1530-8bc7-418a-8ac9-ff191d2fee0b@f63g2000hsf.googlegroups.com>


On Mar 17, 9:32 pm, Frank van Bortel <frank.van.bor..._at_gmail.com> wrote:

> Warning: ROT ! Warning: ROT!
>
> Try to change "large" table to parallel. You will
> need to find out what large means in your case; I've
> never seen anything benefit from parallel queries when
> less than a couple of hundred thousand records need
> to be processes.
> Anything short of that, and the overhead on parallel
> processes is more than the benefit.
> You may also want to play with the number of parallel
> processes
>
> Again: Rule-Of-Thumb warning!
>
> DO NOT take this for granted!

Thanks for your answer Frank.

That was my main idea as well. But there is more to it than that. For instance (and certainly under specific circumstances I haven't tried to determined), a query with an inner join of a huge table and a tiny one won't be parallelized unless the degree of *both* tables is greater than 1. Received on Tue Mar 18 2008 - 03:59:02 CDT

Original text of this message