Re: Io calibration on Exadata

From: Gerald Venzl <gerald.venzl_at_oracle.com>
Date: Wed, 12 Jul 2017 20:09:55 -0700
Message-Id: <E5D99181-C65E-43FD-B644-45BC76FB9A6E_at_oracle.com>



As always with performance there can be many reasons why something got slower and it’s hard to tell without further details.

However, have you made sure that you have set the system statistics to ‘EXADATA’? Gather system statistics in Exadata mode if needed

Priority	Added	Machine Type	OS Type	Exadata Version	Oracle Version
n/a	Auguest 2012	X2-2(4170), X2-2, X2-8, X3-2, X3-8, X4-2	Linux	11.2.x +	11.2.0.2 BP18 and 11.2.0.3 BP8
Benefit / Impact

Gathering Exadata specific system statistics ensure the optimizer is aware of Exadata scan speed. Accurately accounting for the speed of scan operations will ensure the Optimizer chooses an optimal execution plan in a Exadata environment. The following command gathers Exadata specific system statistics

exec dbms_stats.gather_system_stats('EXADATA');

Note this best practice is not a general recommendation to gather system statistics in Exadata mode for all Exadata environments. For existing customers who have acceptable performance with their current execution plans, do not gather system statistics in Exadata mode. For existing customers whose cardinality estimates are accurate, but suffer from the optimizer over estimating the cost of a full table scan where the full scan performs better, then gather system statistics in Exadata mode. For new applications where the impact can be assessed from the beginning, and dealt with easily if there is a problem, gather system statistics in Exadata mode.

Found in, and you may want to review: Oracle Sun Database Machine Setup/Configuration Best Practices (Doc ID 1274318.1)

Thx,


Gerald Venzl | Senior Principal Product Manager Email: gerald.venzl_at_oracle.com <mailto:gerald.venzl_at_oracle.com> | Phone: +1.650.633.0085 <tel:+16506330085> Oracle ST & Database Development
400 Oracle Parkway | Redwood Shores | 94065 | USA

> On Jul 12, 2017, at 19:20, dedba_at_tpg.com.au 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

> --
> Sent from my Android device with K-9 Mail. Please excuse my brevity.

--
http://www.freelists.org/webpage/oracle-l
Received on Thu Jul 13 2017 - 05:09:55 CEST

Original text of this message