Re: DBWR - How Many is Too Many?

From: Nuno Souto <dbvision_at_iinet.net.au>
Date: Fri, 29 Feb 2008 00:38:04 +1100
Message-ID: <47C6B93C.1030307@iinet.net.au>


David Barbour wrote,on my timestamp of 28/02/2008 5:53 AM:
> We recently moved our database to a new SAN. Performance has just
> tanked. Here's the environment:
> AIX5.3L
> Oracle 9.2.0.7 <http://9.2.0.7>
> SAN - IBM DS4800
Aye...

> sar is scary (just a small portion)
>
> AIX r3prdci1 3 5 00CE0B8A4C00 02/27/08
>
> System configuration: lcpu=8
>
> 00:00:00 %usr %sys %wio %idle physc
> 02:15:01 19 19 42 19 4.00
> 02:20:00 21 25 40 14 4.00
> 02:25:00 19 18 43 20 4.00
> 02:30:00 18 18 43 21 4.00
> 02:35:00 20 24 40 16 4.00

wio is w-a-a-a-a-a-y too high with sys time low, might well be a I/O misconfiguration.

> 1. One db writer for each database disk - in our case that would be 8
> 2. CPUs/8 adjusted for multiples of CPU groups - in our case that would
> be 4
>
> Any thoughts?

Large datablock waits in db, plus lots of wio in the OS.

I'm thinking possible misconfiguration
of the AIX Disk I/O pacing parameters.
 From memory, it's:
smitty chgsys
and check all file systems for minpout and maxpout values. Unless you know in detail the I/O balance of your system, you're much better off with these two set at 0 (zero).

There are papers at IBM on this subject, gogle for "IBM AIX disk I/O Pacing".

Had a similar situation a while ago: a new SAN's file systems for some reason ended up with 30/30 for both pout pars. When we set them to 0, the wio dropped very significantly.

At least worth checking.

-- 
Cheers
Nuno Souto
in rainy Sydney, Australia
dbvision_at_iinet.net.au
--
http://www.freelists.org/webpage/oracle-l
Received on Thu Feb 28 2008 - 07:38:04 CST

Original text of this message