How to shrink SGA pools with large granule size and subpools?
Date: Wed, 13 Aug 2014 11:04:51 -0500 (CDT)
Message-ID: <d76c841485fc325d75f49bef25a9a23f.squirrel_at_society.servebeer.com>
Hey all,
In 11.2.0.3 on AIX 7.1, I noticed that my new Prod instance has 2GB allocated to both JAVA_POOL_SIZE and LARGE_POOL_SIZE. I've manually sized the SGA (no sga_target or memory_target). I'm not using (much) Java and had set the large pool according to RMAN usage calculations, which comes to 176MB.
Oh yeah, I forgot about granule size! Hmm, with the buffer cache set to 128GB, it looks like the 2GB shared pool pushes the granule to 512MB. And I see that the subpool size is set to 4 for our 16-core box, so the 2GB allocation now makes sense to me (512MB * 4 subpools).
The extra 3GB of RAM in these pools seems wasteful to me. Has anyone ever gone through the exercise of reducing them? Any problems or side-effects? I could just drop the buffer cache by 3GB or so, I suppose, but I'm also considering the future.
Rich
-- http://www.freelists.org/webpage/oracle-lReceived on Wed Aug 13 2014 - 18:04:51 CEST