Re: Io calibration on Exadata
- From: Mladen Gogala <gogala.mladen@xxxxxxxxx>
- To: dedba@xxxxxxxxxx
- Date: Thu, 13 Jul 2017 00:13:59 -0400
On Thu, 13 Jul 2017 12:20:56 +1000
dedba@xxxxxxxxxx wrote:
Hi,
I've migrated a datawarehouse to my new exa and now ran io_calibrate on it.
To my horror it returns worse figures than i expect, eg 500mb per second
throughput. The old & tired ODA clocks in at 3 times that, and more.. We do
see good performance on real queries, so the calibration is cearly wrong..
Any one has any pointers? Should I use auto-dop at all on an exadata?
Cheers,
Tony
That depends on the release of your database. In Oracle 12.2, I would say yes,
without a hesitation. In Oracle 11G, the answer would probably be no. A good
article on CALIBRATE_IO can be found here:
https://www.pythian.com/blog/secrets-of-oracles-automatic-degree-of-parallelism/
Other than that, I wouldn't necessarily trust CALIBRATE_IO unconditionally.
Testing your IO is a complex task which requires a bit of planning. I would
also run Orion and SLOB to verify that the number produced by CALIBRATE_IO are
accurate. Here is another must-read article about CALIBRATE_IO:
https://kevinclosson.net/2016/06/02/yes-you-must-use-calibrate_io-no-you-mustnt-use-it-to-test-storage-performance/
And finally, I have to point you to the Mauro Pagano's wonderful explanation of
AutoDOP:
https://mauro-pagano.com/2015/07/29/autodop-in-12c-whats-new-and-whats-old/
If the results seem wrong, feel free to correct them by updating the table
SYS.RESOURCE_IO_CALIBRATE$. Long story short, I've heard horror stories about
CALIBRATE_IO in 11G and I've heard satisfied comments in 12c. Not being a DBA,
I don't have any relevant experience of my own.
Regards
--
Mladen Gogala
Oracle DBA
Tel: (347) 321-1217
--
//www.freelists.org/webpage/oracle-l
Other related posts: