Ram,
Not sure there is a solution as such to direct reads because it's deliberate
behaviour.
When direct reads are a problem it's often just inefficient execution plan -
either a flip to a bad plan or just previous inefficiencies being hidden by the
buffer cache?
So if it's a problem then do fewer full segment scans? Tune queries...
Otherwise you can influence the inputs into the adaptive decision - there are
quite a few articles out there about this.
E.g
Larger buffer cache
Fiddle with the thresholds
Etc
Regarding changes in 12c, this is not something I've investigated personally.
Nor do we have any prod systems on 12c within my scope at work.
Mladen said earlier that he had an impression it was less of an issue.
Maybe?
Big table cache perhaps?
http://progeeking.com/2014/10/03/rac-12c-direct-path-reads/
Hope this helps.
Dom.
Sent from my Windows Phone
________________________________
From: Ram Raman<mailto:veeeraman@xxxxxxxxx>
Sent: 22/10/2015 03:47
To: dombrooks@xxxxxxxxxxx<mailto:dombrooks@xxxxxxxxxxx>
Cc: gogala.mladen@xxxxxxxxx<mailto:gogala.mladen@xxxxxxxxx>;
ORACLE-L<mailto:oracle-l@xxxxxxxxxxxxx>
Subject: Re: Database perfortmance degraded drastically after upgrading the
database from Oracle 102.0.4 to Oracle 11.2.0.3
For direct path reads, it's worth it. It's one of the most common cause of
problems on this upgrade path.
