Nice direct measurement.
Another method is to have a re-executable regression test and measure the polar
cases nothing pinned versus everything pinned and the incremental n cases of
selected pinning and simply report elapsed time and maximum shared pool
occupied (or just elapsed time if the latter is too tricky in the environment).
I have not done this in quite a while, but I was a bit surprised at how often
“pin everything” won and by how much.
Nice work Tanel!
From: oracle-l-bounce@xxxxxxxxxxxxx [mailto:oracle-l-bounce@xxxxxxxxxxxxx] On ;
Behalf Of Tanel Poder
Sent: Monday, February 04, 2019 5:19 PM
To: Jonathan Lewis
Cc: Oracle-L Freelists; martin.klier@xxxxxxxxxxxxxxxxx
Subject: Re: ASMM - resizing triggers/thresholds
Hi,
In addition to what Jonathan said, here are my ramblings about what the KGL
simulator is (and the buffer cache simulator uses similar logic) that is used
for modeling "time saved reloading objects/blocks into cache if we hadn't
kicked them out previously":
https://blog.tanelpoder.com/2009/09/13/kgl-simulator-shared-pool-simulator-and-buffer-cache-simulator-what-are-these/
--
Tanel Poder
https://blog.tanelpoder.com/seminar/
On Sat, Feb 2, 2019 at 9:05 AM Jonathan Lewis <jonathan@xxxxxxxxxxxxxxxxxx>
wrote:
Martin,
If you look for the hidden parameters like "%memory_broker%" that gives you
some clues.
12.2.0.1
NAME SES_VAL
--------------------------------------------- ------------------------------
_automemory_broker_interval 3
_memory_broker_log_stat_entries 5
_memory_broker_marginal_utility_bc 12