It’s such a mess with those histograms and statistics that I had to rewrite this post at least ten times. It’s such a mess that I still cannot believe how poorly maintained the global statistics are and how badly created the histograms are. Because of this mess and the depth of the topic this post is a bit longer than usual, but I find it really useful and I think it can give answers to many questions.
Next March 19th I’ll be speaking at OUG Ireland in Dublin, presenting some cool stuff around Oracle Database 12c, PDB Cloning, ZFS snapshots and Change management through EM12c. The topic is easy going and can be interesting for DBA’s and developers as well. The abstract and agenda of OUG Ireland is in here http://bit.ly/1vg3Lie .
See you soon.
Oracle is full of surprises, dbms_stats and optimizer statistics as well. That’s why I’m starting a new series named “Optimized statistics” in which I’ll cover some gathering and usage issues related to them.
As we know 11g has a cool feature which tracks how much data we had changed in our objects and if it’s more than the 10% /default/ it marks our statistic as stale. This is really useful because we don’t have to waste time scanning objects without having changed reasonable amount of data in them. But is it working for partitions? Well, not really, not till today.
Yes, direct path reads again :)
No worries I’m already a bit bored from digging in this algorithm, so I think this is going to be my last post about it. Till they change it again of course. And yes, in 220.127.116.11 they did.
In Oracle Database 12c we can find many new and shiny things… So many that we can miss the little good things really easy. I think this one, is one of them.
As simple as that!
A fast and straight approach for creating local IPS repository.