Thoughts on Parallelization

From: Don Seiler <don_at_seiler.us>
Date: Tue, 15 May 2007 15:48:13 -0500
Message-ID: <716f7a630705151348l5522261creda27cc18f8cabf1@mail.gmail.com>

I've been doing some reading on parallelization, after seeing Doug Burns' presentation on the Pythian blog a while back. In it, Doug mentioned that he doesn't set any degree of parallelization on tables or indexes, and relies on query hints to use parallel query.

I'm wondering what others might think of this, or if relying on Oracle's multi-user adaptive logic to control things is fine. I realize that there might be some extreme cases, but I'm wondering if you people are happily using the default parallelism on your tables and indexes and living life to the fullest.

Also, what about the equation for setting parallel_max_servers. The documentation suggests:

CPU_COUNT x PARALLEL_THREADS_PER_CPU x (2 if PGA_AGGREGATE_TARGET > 0; otherwise 1) x 5

My CPU_COUNT is 8, but that is due to linux hyperthreading on 4 physical CPUs. So even then I'm looking at 4 x 2 x 2 x 5 = 80 paralllel_max_servers. (Running 10.2.0.2 on RHEL3). I'm not sure what to expect so I'd like to see what others are using in their setups, if you're willing to share.

--

Don Seiler
oracle blog: http://ora.seiler.us
ultimate: http://www.mufc.us
--

http://www.freelists.org/webpage/oracle-l Received on Tue May 15 2007 - 15:48:13 CDT