RE: Why there are lots of page-in and almost no page-out?
- From: "Gogala, Mladen" <Mladen.Gogala@xxxxxxxx>
- To: "'roger_xu@xxxxxxxxxxx'" <roger_xu@xxxxxxxxxxx>, "Oracle-L@Freelists. Org (E-mail)" <oracle-l@xxxxxxxxxxxxx>
- Date: Tue, 14 Sep 2004 14:44:46 -0400
It's hard to tell without getting my hands on your system, but page in
is mapping a new page in the address space of a process. In particular,
if a process needs new stack pages, those will be allocated from
the "free pages" (pages that have a valid backup on the disk), and handed
to the process which requested free pages. That mapping will be recorded as
a "page in". If you have sufficient free memory, so that paged daemon is not
activated, you will not see any page-outs. There are two types of page ins,
girly page-ins, when pages are reclaimed from memory and placed in the
address space of a process, and manly page-ins, when page has to be brought
in from the disk. In non-arnoldian terminology, it's soft page faults and
hard page faults. Address space manipulations happen while CPU is in kernel
mode. If your top monitor (free version of Glance Plus, much inferior to it)
shows that you are spending more then 10% in kernel mode, you have a
problem.
--
Mladen Gogala
A & E TV Network
Ext. 1216
> -----Original Message-----
> From: Roger Xu [mailto:roger_xu@xxxxxxxxxxx]
> Sent: Tuesday, September 14, 2004 2:26 PM
> To: Oracle-L@Freelists. Org (E-mail)
> Subject: Why there are lots of page-in and almost no page-out?
>
>
> Hi,
>
> We are running Oracle 9.2.0.4.0 on Solaris 9 and please see
> the output from "vmstat". Do you think we have a paging
> problem? We are using plain UFS and I do not think direct/IO
> is turn on.
>
--
//www.freelists.org/webpage/oracle-l
Other related posts:
