Re: Process memory behaviour of Oracle
Date: Mon, 14 Jan 2008 10:41:44 -0800 (PST)
On Jan 14, 6:21 pm, "Jonathan Lewis" <jonat..._at_jlcomp.demon.co.uk> wrote:
> The presentation and notes at the following URL may help:http://www.pythian.com/blogs/741/pythian-goodies-free-memory-swap-ora...
> Jonathan Lewishttp://jonathanlewis.wordpress.com
In the meantime I have also noted the following: when I check v $pgastats I notice a huge entry for memory freed/released to OS (suspiciously close to the 10G I released earlier) - I am guessing that Oracle "holds on" to PGA memory for a while and does not let the OS have it back ?
From what I understand of the Linux memory allocator it masks system brk() calls inside the library malloc() for efficiency reaons so perhaps Oracle does something similar ? But one problem I have is that the SGA sometimes has problems acquiring memory (I see this in the stack trace of a Java OJDBC driver when it tries to set up connections etc.) even though my own C program does not (i.e., I believe the memory is there to be had)
...but first I must study your link! Received on Mon Jan 14 2008 - 12:41:44 CST