Re: CPU LOAD on hp T500: ORACLE 7.1.4

From: Steve Adams <steveadams_at_acslink.net.au>
Date: 1998/01/06
Message-ID: <34b2bbce.1986951296@newsserver.trl.oz.au>#1/1

Hi Bruno,

This has the hallmarks of standard SMP memory contention because of there being too many active CPUs and not a high enough memory interleave factor. It is a well known problem on high end T series boxes. Often you can improve your memory interleave factor by changing the memory config. Thereafter switching to raw datafiles to reduce the load on the Unix buffer cache is the most productive strategy. Also, you should try to prevent filesystem traversals (finds, backups and so on) during production hours. Also avoid file and directory names longer than 14 chars each, minimise NFS activity and consider runing syncer at say 60 seconds instead of the default of 30 seconds. You can also mount your file systems with the write behind option, and of course any tuning of the I/O subsystem configuration will help.

Hope this helps,
Regards,
Steve Adams
steveadams_at_asclink.net.au

---

On 6 Jan 1998 14:03:52 GMT, "menon bruno" <bruno.menon_at_francetelecom.fr> wrote:

>hello,
>
>I have a very great problem on my system.
>
>server : hp9000 / T500 => 7 Proc = 1,5 Go RAM
>os : hp-ux 9.04
>SGBDR : ORACLE 7.3 whith 5 DB interconect by sqlnet V2 on the server
>
>user arrive by telnet
>
>
>since some day the system has many problem.
>
>the fact is = > cpuload
>on vmstat :
>cpu usr = 0%
>cpu idle = 0%
>cpu sys = 100 %
>
>the process's consumming cpu are :
>oracleDB LOCAL=NO
>this some sqlnet process between DB.
>
>
>Is anyone have the same problèm in the past ?
>
>Thank you.
>
Received on Tue Jan 06 1998 - 00:00:00 CST