Oracle FAQ | Your Portal to the Oracle Knowledge Grid |
Home -> Community -> Mailing Lists -> Oracle-L -> Re: linux memory: limiting filesystem caching
Zhu,
You wont be able to see the memory benefits by just monitoring free memory.
"System hung"
I over used the statement. I was refering to a system which is so slow
that it takes .. say 45 seconds to connect to Oracle.
"I don't understand this. Every session, would access the whole SGA on its session creation. right?"
No, it only touches a small part of it on connect. Unless you set pre-page SGA, which i dont recommend. You can see this with top and the RSS memory column. Connect, check the value of your shadow process, run a high gets query, and check RSS again.
Most kernel parameters help in very specific tasks, and will only be seen via a specific test.
Christo
On 7/13/05, zhu chao <zhuchao_at_gmail.com> wrote:
> Hi,
> Actually system has been down for almost one year and I have lost
> the statistics. I remember I checked SAR log to compare user%, SYS%
> CPU usage. I can't find significant difference.
> As for memory, I check free log (I have cronjob to report memory
> usage to log file). My system used middleware and we don't have often
> connect-disconnect behavior. I don't have system hung (once actually
> during the 2 year's this RAC system lifecycle, unknown reason, 7445
> and system panic). We don't do heavy filesystem IO operation. I have
> 1.7GB sga and 4GB physical memory.
>
> >and if your sessions
> > are touching the entire SGA,
>
> I don't understand this. Every session, would access the whole SGA on
> its session creation. right?
>
> I was the very early user of oracle 9.2 RAC user and did play some
> so-called tuning at OS level, painfully I did not see much performance
> gain, including those network, kernel parameter tuning. One
> significant thing is max open file must be increased else when session
> connections high, oracle unable to open file.
>
> On 7/13/05, Christo Kutrovsky <kutrovsky.oracle_at_gmail.com> wrote:
> > Zhu,
> >
> > How did you measure the performance differences when using bigpages
> > and not using them? What kind of tests did you run ? Were they CPU or
> > disk IO specific ? Do you have statistics of CPU usage in "user" mode
> > and "system" mode?
> >
> > We have at least 3 cases where converting a production system to
> > bigpages/hugepages was the difference between a hung system and a well
> > performing system.
> >
> > As for memory savings, what counter-arguments do you have? Keep in
> > mind that PTE pointers are allocated as needed, and if your sessions
> > are touching the entire SGA, for example as often happens, they
> > reconnect often.
> >
> > I agree that it doesn't sound reasonable, thus the hugepages solution.
> >
> >
> > --
> > Christo Kutrovsky
> > Database/System Administrator
> > The Pythian Group
> --
> Regards
> Zhu Chao
> www.cnoug.org
>
-- Christo Kutrovsky Database/System Administrator The Pythian Group -- http://www.freelists.org/webpage/oracle-lReceived on Wed Jul 13 2005 - 10:17:24 CDT