Experience with parallel query option
Date: 1995/04/11
Message-ID: <3me5la$sbh_at_caladan.restena.lu>#1/1
I have played around a little bit with the parallel query option of ORACLE
7.1.4.
I found out that for short queries there is a performance decrease when using
the parallel queries. This seems logical as there must be some overhead for
splitting the query over several processors.
The query I used was very simple:
select count(*) from table,table;
I used table,table in order to get the cartesian product. This allowed me to have a big result set without having to have very big tables. I executed this query and then I gradually incremented the number of rows in table. In SQL*PLUS with "set timing on" I measured the times once without parallel once with parallel (degree 2 as my server has 2 processors). To change to parallel I used "alter table table parallel (degree 2)".
As I said before at the beginning the queries were faster without the parallel query option. From the approx. count(*)=12000000 on, the parallel query started to be faster than the normal one.
Questions:
- Does this suggest that the parallel option is only useful for huge queries (very many rows or long running queries)?
- Is it useful to set the degree to a value higher than the number of processors on the server?
- Has anyone made other experiences with this option?
- Was this the correct way of testing the option?
Thanx for any comments Received on Tue Apr 11 1995 - 00:00:00 CEST