RE: ASMM
Date: Fri, 16 May 2014 16:50:08 -0400
Message-ID: <208201cf7148$6d019d00$4704d700$_at_rsiz.com>
That is a possibility. Looking at your resize operations will tell you what is going on.
Two bad cases are possible: 1) lots of literals and/or unique queries so the shared pool wants to grow and AMM pushes the boundary; 2) Hourly grow shrink near the stable point swapping only a few granules between shared pool and buffer cache, so both recent buffer reads and recent parses tend to get discarded.
Both are good reasons to turn AMM off, especially if you know about what a good size range is for the overall SGA.
From: oracle-l-bounce_at_freelists.org [mailto:oracle-l-bounce_at_freelists.org] On Behalf Of Dragutin Jastrebic
Sent: Friday, May 16, 2014 3:59 PM
To: Brian.Zelli_at_roswellpark.org
Cc: oracle-l (oracle-l_at_freelists.org)
Subject: Re: ASMM
=>
I am finding that my Library Cache Get hit Ratio % is hovering around 50-60
percent. I am running AMM and I set
the memory_target twice what it was and still no relief…..
Brian
<=
Hello
I remember that I have faced a strange behaviour with 10g and SGA_target on one or even two projects. Oracle was constantly resizing the shared_pool and the buffer_cache and this shrink-grow activity between the two was constantly eliminating some queries from the shared_pool, lowering the library cache hit ratio. The view that shows this shrink-grow behaviour inside the SGA is v$sga_resize_ops. It was couple of years ago so hopefully this behaviour is improved with new patches/versions of Oracle, however maybe you can still check this.
HTH Dragutin
-- http://www.freelists.org/webpage/oracle-lReceived on Fri May 16 2014 - 22:50:08 CEST