Oracle FAQ Your Portal to the Oracle Knowledge Grid
HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US
 

Home -> Community -> Usenet -> c.d.o.server -> Re: ORACLE on Linux - IO bottleneck

Re: ORACLE on Linux - IO bottleneck

From: Noons <wizofoz2k_at_yahoo.com.au>
Date: 8 Feb 2006 18:05:51 -0800
Message-ID: <1139450751.368884.140590@o13g2000cwo.googlegroups.com>


Wyvern wrote:
> Hello,
>
> we've RedHat Linux AS3 (upd. 1) running in a 8 proccessor IA64 Itanium
> II machine with RAM-16Gb. Oracle 9i (OLTP database) under RawDevices
> with
> an oracle blocksize of 8k and 2 QLogic Fibre Channel Adapters (QLA2340)
> with
> Symmetrix EMC disk array. We also have ASync I/O activated in oracle.
> Kernel version is "2.4.21-9.EL #1 SMP"

Hmmmm.....
You have a 64-bit processor system, but the version of RH you're running seems to be the 32-bit version and then again four upgrade levels below what Oracle recommends as a minimum (upd.5 instead of upd.1) for 9i.

I'd first look into getting the correct version of Linux: 64-bit and definitely higher than RHAS3-upd.1.
And then a 64-bit version of Oracle for Linux. Call Oracle support and get them to tell you the exact versions of everything you should be running: I' ve found that's faster than pouring through all the thousands of conflicting, out of date and out of context information in as many sites.

Only after that would I start looking into IO problems.

> We've different DDBB with similar characteristics running under similar
> hardware but with EXT3 and we've got much better performance.
> We also have RAW DEVICES under AIX (4.3 and 5.2) and the performance is

Have you patched up Oracle and RH with all the required fixes to run aio in 9i and RH3? There's a bunch of them for both. The 64-bit thing is another clanger in that equation.

If I were you, I'd be looking at running at the very least RHAS4 for 64-bit with Oracle 9i for 64-bit. RHAS4 seems to require a lot less patches to get going with Oracle and aio.

Having said that: we're running RHAS3-upd.5, 9.2.0.6, all patched up, and just direct IO. 8K block size. Without a saturated IO controller, I'm getting more than 500 direct IO/sec operations in a test program. But when the controller saturates, I get a lot less than that. It's a good
indicator of when we need to start splitting things across controllers/hyperchannels. Received on Wed Feb 08 2006 - 20:05:51 CST

Original text of this message

HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US