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: Fabrizio Magni <fabrizio.magni_at_mycontinent.com>
Date: Fri, 10 Feb 2006 18:51:50 +0100
Message-ID: <43ecd2b5$0$28071$4fafbaef@reader1.news.tin.it>


Mladen Gogala wrote:
> On Thu, 09 Feb 2006 15:14:20 +0100, Fabrizio Magni wrote:
>

>> The second dd has its I/O merged.
>>
>> Consider this controversial but the I/O scheduler eliminate the need to 
>> have oracle block size "aligned" to file system block size (even without 
>> direct I/O).
>>

>
> How exactly did you test? Did you use elvtune (RH3) or did you put an
> appropriate scheduler in grub.conf (RH4)? This is a very special case of
> a sequential I/O which doesn't tell you much. Usually, I/O patterns
> resulting from the database use are a bit different.
>

Test was performed on a SLES9 with cfq scheduler. The result would be the same even with the other three (or with the older elevator).

A dd doesn't represent the typical database workload but it shows what an I/O scheduler can do.
This component relax the rule of thumb "db block size=file system blocksize" suggested by many.

For example: with a db block size of 8k and a filesystem of 4k every oracle request isn't translated in two I/O of 4k each but in a single 8k operation.
The I/O merging is only one of the function of the scheduler.

If you play with ETL and datawarehouses you can have significant performance enhancement.

For an OLTP the deadline scheduler could be a good solution.

-- 
Fabrizio Magni

fabrizio.magni_at_mycontinent.com

replace mycontinent with europe
Received on Fri Feb 10 2006 - 11:51:50 CST

Original text of this message

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