Re: Details needed on using the parallel query option

Brian M. Biggs wrote:
> The way I understand the parallel query option, it can be invoked
> primarily in 2 ways:
> (1) by ALTERing a table and indicating that it is a parallel table
> (2) putting hints in your SQL code to indicate that a certain SQL
> statement should be executed in parallel.-SNIP-
> Is there a way to tell Oracle "only execute this query in parallel if
> you have more than 1 processor"? Or, rather than being that specific,
> is there a better way to control when your application does and does not
> use the parallel query option? What is the best way to manage this
> feature?

What you need is a PRO*C program that calls the sysconf() system call. With the "_SC_NPROCESSORS_CONF" argument it will return the number of cpu's (processors) the system has. Then the PRO*C program could do the proper "alter table schema.table_name parallel(degree n)" or "alter table schema.table_name noparallel" based on the tables overall size as returned with a query against dba_segments.

Looking at the man pages for sysconf() on AIX 3.2.5 I see it doesn't return the number of cpu's (assume 1), AIX 4.1 should work fine, as does SUN Solaris 2+. I didn't see anything on our HP9 and we don't have a HP10 so I don't know what to do in the HP environment.

Also, don't forget the required init.ora parameters for parallel query.

-- 
                       \\|//
                       (0-0)
           +-----oOO----(_)-----------+
           | Brian P. Mac Lean        |
           | Database Analyst         |
           | brian.maclean_at_teldta.com |
           | http://www.teldta.com    |
           +-------------------oOO----+
                      |__|__|
                       || ||
                      ooO Ooo