Re: Parallel Query

We need a few more clues about what
you are doing.

Couple of thoughts though -
Number of discs may have a big impact,
sort_area_size and location of temp
tablespace may have a big effect.

As an example, if you are doing:

    select /*+ parallel (t , 4) */ col1, sum(col2)     from big_table t
    where col3 = 'xxx'
    group by col1

Where the condition returns 1M rows, and col1 has a couple of dozen values.

You could get close to a factor of 4 improvement on just 2 CPUs if you can avoid I/O contention.

--

Jonathan Lewis
Yet another Oracle-related web site:  http://www.jlcomp.demon.co.uk

Dean wrote in message <8kiblt$imq$1_at_uranium.btinternet.com>...


>Hi,

>Has anyone out there had any succes using Parallel Query?

>If so, can you give me a idea of the improvements seen, size of the tables

>(rows) and kit the RDBMS was running on.

>

>I'm struggling to get improvment on a 13Mill row table on a NT box with

>2x800mghz processors.

>I'm going to try with a 4x200mghz NT and a 4 processor IBM UNIX box but

>would be grateful for any pointers first.

>

>Many thanks in advance

>

>Dean

>

>