RE: Oracle Read Consistent Overhead
- From: Matt McClernon <mccmx@xxxxxxxxxxx>
- To: <jonathan@xxxxxxxxxxxxxxxxxx>, <oracle-l@xxxxxxxxxxxxx>
- Date: Wed, 20 Apr 2011 00:12:09 +0000
> Read-consistency costs are largely about the number of undo records applied,
> not
> about the number of blocks in the underlying object, and the number of undo
> records is (generally) related to the number of changes, which often means
> number of rows.
Ignore my last post. I re-ran the test under more controlled (isolated)
conditions and the CR overhead is identical with and without the index on the
MV log. The ratio for both is 2 CRs per row update which seems perfectly
acceptable based on your feedback that consistent reads are related to row
changes rather than blocks changes.
My first test results must have been polluted by another session/job.
I did notice that the MV refresh process redundantly updates the SNAPTIME$$
column for the rows which it propagates to the MV before proceeding to delete
those same rows. I guess this must be a side effect of Oracle using the same
code path for both 'on demand' and 'on commit' refresh MVs.
Thanks very much for your help
Other related posts:
