One design issue about merge and equivalent DDL

Question

Hi All,
My DB is 11.1. RAC 4 nodes.
Today I encountered one SQL that did merge and ran very slowly. After tuning, it
cost more than 2 hrs to finish.
According to Oracle manual and my own understanding,
1.MERGE itself can't have inner-partition parallelism. So if the target table is
non-partitioned, PML doesn't work here.
2.MERGE genertes redo and can't be suppressed.
So I think whether an equivalent CTAS works better, considering about daily
increments are relatively large amount comparing to the target table.
Pseudo SQL looks like:
create table new_table
parallel 4 nologging
as
select case
when target_table_join_columns is null
then new_line
when source_table_join_columns is null
then original_line
when both are not null
then updated_line
from target_table t full outer join source_table s
on ... join columns;
(It's just a sample.)
The advantages I can see is:
1. Far less redo.
2. Parallel
The disadvantages I can see is:
1. More spaces
2. Full load not incremental.
I know I have to benchmark it and consider about all cases. But has anybody encountered the similar issues before? And whether you get benefits from this kind of change?
Best regards,
Leon

Answer 1

Hi,
you have to set the correct permissions on the HFMAnswers Folder.
But it depends a bit on your exact requirement....users must be able to save reports under the shared folder or only my folders?
What you need a least are some groups to make distinction between who can see which reports and who can use which functionality within oracle bi.
Do you have that clear? This way you can try to setup a generic security framework.
Kr,
A