Richard,
Excellent testing approach! Thanks so much! I'll try it...
-Tim
on 10/3/03 6:50 AM, Richard Foote at [EMAIL PROTECTED] wrote:
> Hi Tim,
>
>
>
> Good questions.
>
>
>
> As you mention, the memory Oracle "says and thinks" it's released and what
> it "actually" releases to the ker
Hi Tim,
Good questions.
As you mention, the memory Oracle "says and thinks" it's released and what
it "actually" releases to the kernel has generally been two different
things. However, the behaviour with P_A_T is somewhat different. A simple
little test for the unconvinced is to simply issue (th
Hi Tim,
Good questions.
As you mention, the memory Oracle "says and thinks" it's released and what
it "actually" releases to the kernel has generally been two different
things. However, the behaviour with P_A_T is somewhat different. A simple
little test for the unconvinced is to simply issue
I remember from somewhere in Ixora, that Oracle does only free() without
calling brk() with a negative value to actually release the memory (probably
you even can't do it in every circumstance), so the memory will remain used
untill process exits or dies.
OTOH, Oracle server processes are meant fo
Answer inline.
On Wed, 2003-10-01 at 13:44, Tim Gorman wrote:
> Richard,
>
> Thanks for the detailed explanation!
>
> As a "C" programmer of some 20 years, I can only assume that Oracle has done
> away with the use of the "malloc()", "free()", etc UNIX library calls and is
> now calling the UNIX
Richard,
Thanks for the detailed explanation!
As a "C" programmer of some 20 years, I can only assume that Oracle has done
away with the use of the "malloc()", "free()", etc UNIX library calls and is
now calling the UNIX system call "brk()" directly?
It was the underlying heap-extent management
Hi Tim,
I would suggest there are two key advantages to using automatic workspace
management.
The first and perhaps most important is that yes, unlike the manual method
by which sessions "cling" onto memory, automatic workspace management can
deallocate the tuneable portion of the PGAs (those pre
Richard,
I take it that your two points are...shall we say...enhancement requests,
not current functionality? :-)
Following up on the discussion of "space-efficiency" and tabling (for the
moment) my questions about the "performance-efficiency" side of things.
Yes, there certainly is an element o
located
memory? This is a real question.
Niall
But I like the conversation idea anyway
> -Original Message-
> From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On
> Behalf Of Richard Foote
> Sent: 29 September 2003 14:30
> To: Multiple recipients of list ORACLE-L
>
Hi Tim,
There are couple of parts of the conversation we've missed out ;)
Firstly, the server process when talking to the P_A_T instance should have
said, "What the hell is going on here, what do you mean I can't have my full
100M, this keeps on happening and it's just good enough. Get a bloody D
Hi Tim,
There are couple of parts of the conversation we've missed out ;)
Firstly, the server process when talking to the P_A_T instance should have
said, "What the hell is going on here, what do you mean I can't have my full
100M, this keeps on happening and it's just good enough. Get a bloody D
> I set PGA_AGGREGATE_TARGET to 2,5G in my home computer, but only about
3-4MB
> of memory was used for my large sort for example. There are parameters
> _smm_max_size and _smm_min_size for setting boundaries for automatic
One more note on these two parameters -> setting them didn't work in my
9.2
Hi!
> From what I've been able to determine about this functionality,
"efficient"
> merely means "space-efficient", not "performance-efficient" (i.e. Fewer
> cycles? Smarter cycles?). Is this correct? Does anyone know of anything
> in WORKAREA_SIZE_POLICY=AUTO which improves performance over
>
13 matches
Mail list logo