Oracle FAQ | Your Portal to the Oracle Knowledge Grid |
![]() |
![]() |
Home -> Community -> Usenet -> c.d.o.server -> Oracle 8i on RS/6000 AIX question.
Thanks everyone for responses!
We are using IBM SSA arrays for our data storage. So far, I am seeing
little to no disk waits occurring.
We have "temporarily" added two processors to the machine (total four
340Mhz RISC III processors) which has drastically dropped my run queue
during heavy periods from an average of about 16 to an average of about
4!
Even with this configuration, the only disk waits I can see are very
short during log switches.
The $64,000 question (which is more than I make!) is, would two 600 MHz RISC IV processors give us the same results? According to IBM and their ROLTP benchmark numbers, it should be even better (57.1 vs. 69). I am wondering if this is due to threads required to manage four processors instead of two.
Just wondered what everyone else thought.
P.S. Sorry about the HTML. This one should be plain.
Todd Parnell
-----Original Message-----
From: chao_ping [mailto:chao_ping] On Behalf Of chao_ping
Posted At: Wednesday, May 01, 2002 2:15 AM
Posted To: server
Conversation: Oracle 8i on RS/6000 AIX question.
Subject: Re: Oracle 8i on RS/6000 AIX question.
Jayaraman Ashok wrote:
> Hi In case you are CPU bound at times, if there are more cpu waits,
then
> i think that if you upgrade the number of cpus, and your disk
> controllers can not match that activity of reads/writes the problem
> would be compounded rather than mitigated. Check your disk i/o
activity
> with the available controllers before decideing to upgrade. Also i
> believe that MTS (multi threaded server) helps alleviate the
problems
> with the memory rather than cpu activity. Also in AIX(RS/6000-Aix
4.3.3)
> the asynchronous processes do consume more of cpu time on behalf of
> their parent oracle processes. Kindly correct me in my above
suggestions
> if they are wrong.
> Regards, Ashok
I agree with ashok. If your datafiles are on jfs filesystem, io consumes a lot of cpus.In my experience, ibm is power in cpu and weak in disk io. Use iostat or vmstat to check the cpu time spent on iowait, maybe helpful. You can also run topas to get an overview of overall system performance bottleneck.
-- an oracle fan, an oracle beginner an oracle fan, an oracle beginner Posted via dBforums http://dbforums.comReceived on Wed May 01 2002 - 07:38:22 CDT
![]() |
![]() |