CIO poor read perf with Oragle in comparison of RBRW mount option

[hervedba]

Dear All,

We had the Oracle file systems configured with mount option rbrw, But I read that Concurrent I/O is most faster.

Than I have created a Oracle file system with CIO, But the read performance are two time slower than RBRW.

I don't understand why??? because RBRW doesn't make file caching (release behind read and write) only the overhead from file caching. And CIO has no overhead for the file caching mecanismus, CIO make direct concurrent read and write.

Have you any idee, why RBRW is faster for the reads than CIO ?


For Information: the disk writes are two time faster with CIO than RBRW


Thanks for any input & Best regards
Hervé

 

[ereshatezer]

What was the effect on overall I/O wait time after changing to CIO?

 

[JCJP]

Hervé,

Can you resolve this problem with CIO ?

We have performance problems in read and write using AIX V.5.3. and Oracle Database 9.2.2 and 10G.

Thanks for your help, regards,

JCJP

 

[mrn]

Give this a read

http://www-1.ibm.com/servers/aix/whitepapers/db_perf_aix.pdf

Mike

"A foolproof method for sculpting an elephant: first, get a huge block of marble, then you chip away everything that doesn't look like an elephant."

http://www.airport-parking-site.co.uk/

 

[JCJP]

MRN,

We already read that white paper, and apply all the recomendations, but specially the disk read process is very low in the p5 with the aix, an Intel (Two 32bit processor) machine with linux run the same procees three times faster than the new P5 520 with two Power5 procesors.

In the Linux machine the process (Oracle Database), an general ledger balance (Sequential reads in two tables, basically) take 50% CPU and the disk I/O reads shown 15-25KB, in the p5 (Aix) the same process show 1-2% CPU using and 3-4 KB. Disk I/O reads.

Regards,