Oracle FAQ Your Portal to the Oracle Knowledge Grid
HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US
 

Home -> Community -> Mailing Lists -> Oracle-L -> Re: AIX pinned memory

Re: AIX pinned memory

From: Mladen Gogala <mgogala_at_verizon.net>
Date: Wed, 13 Dec 2006 07:54:03 -0500
Message-id: <1166014443l.3579l.0l@medo.noip.com>

On 12/12/2006 05:21:10 PM, Allen, Brandon wrote:
> Howdy listers,
>
> This is actually more of a Unix question, but it is also Oracle related
> and I know there are a lot of people on this list more experienced than
> I with Unix so I figured I'd give it a shot.
>
> I'm running Oracle 10.2.0.2 on AIX 5.3 and am using the pinned SGA
> feature (lock_sga=true & v_pinshm=1). I'm also using the large (16MB)
> pages, but I'm not sure if that is relevant.
>
> The problem is that the output of vmstat -v and svmon -G both show that
> I have about 1.9million pages (1.9M*4K=7.2G) that are currently pinned,
> however my Oracle SGA is only 3G, so I'm concerned about the other 4.2GB
> that appear to be pinned. We aren't having any paging or swapping or
> other performance problems so this is just a curiosity at this point.

Brandon, the answer to your question is clarification what the word "pinned" actually means. When we say that memory is "pinned" it actually means that the page table prohibits page stealing and swapping. In other words "paged" or "pdflush" or "update" or whatever this daemon is called on AIX, can not throw pages from this page table out and replace them with other pages. Swapper also cannot write throw this segment out. What your monitoring programs are showing you is likely the maximum size of the page table, not the actual pinned memory. Of course, that would be my guess. The real people to answer that question would be the support engineers at IBM. It's better manually.

-- 
Mladen Gogala
http://www.mladen-gogala.com

--
http://www.freelists.org/webpage/oracle-l
Received on Wed Dec 13 2006 - 06:54:03 CST

Original text of this message

HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US