Antw: Re: [suse-sles-e] SLES and Oracle Memory Problem?

From: matilda matilda (matilda_at_grandel.de)
Date: Wed Jun 13 2007 - 09:54:50 CEST

Next message: Thorsten Kukuk: "Re: [suse-sles-e] When does SLES 10 SP1 appear?"
Previous message: Matthias Eble: "Re: [suse-sles-e] SLES and Oracle Memory Problem?"
In reply to: Matthias Eble: "Re: [suse-sles-e] SLES and Oracle Memory Problem?"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] [ attachment ]

Message-Id: <466FBEEB.117B.0044.0@grandel.de>
Date: Wed, 13 Jun 2007 09:54:50 +0200
From: "matilda matilda" <matilda@grandel.de>
Subject: Antw: Re: [suse-sles-e] SLES and Oracle Memory Problem?

Hi Matthias, hi Julian, hi all,

In my opinion you have to watch the real swapping activity with
vmstat over time. We do have a system with 8 oracle instances and
6GB real memory and 4GB swap.

We do have 1,5GB swapped out and 3,7GB cached.
The first time I was wondering about that I researched that the
kernel policy can decide to put infrequently used pages to swap
when it assumes that buffer/cache is more valuable.

Our instances are at most mutual exclusive, that means, they are used
at different times or only for testing without huge load.

In y opinion you have to think about:
1) Check the target SGA memory size of your oracle instances.
Check what they really need. (e.g. test database with lower target SGA size).
The sum of all SGA target sizes should not be more than pysical memory - some
amount for other processes.
2) Vmware server reserves the amount of memory for the guest as defined
at the vmguest configuration. You can see this pretty good with top.

Best regards
Andreas Mock

>>> Matthias Eble <matthias.eble@mailing.kaufland-informationssysteme.com> 13.06.2007 09:32 >>>
Julian,

there is a heavy formating problem with your mail with my mail-client. I
suppose free -m looks like this:

output:jason:~ free -m
total used free shared buffers cached
Mem: 3950 3939 11 0 203 3263
-/+ buffers/cache: 473 3477
Swap: 2054 26 2028

I'd say your system is fine since a large amount of memory is used for
caching only and would AFAIK be freed if a process asks for memory.
I think, depending on IO traffic speed could suffer if buffer/cache free
would get scarce.

That's what I think were the results of my researches a couple of years
ago. I'd really like to know if this is definitely true, since these
questions come up in our company very often.

Thanks
Matthias

---------------------------------------------------------------------
To unsubscribe, e-mail: suse-sles-e-unsubscribe@suse.com
For additional commands, e-mail: suse-sles-e-help@suse.com

Next message: Thorsten Kukuk: "Re: [suse-sles-e] When does SLES 10 SP1 appear?"
Previous message: Matthias Eble: "Re: [suse-sles-e] SLES and Oracle Memory Problem?"
In reply to: Matthias Eble: "Re: [suse-sles-e] SLES and Oracle Memory Problem?"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] [ attachment ]

This archive was generated by hypermail 2.1.7 : Wed Jun 13 2007 - 11:59:00 CEST