Printable View
hi,
please click this link to see the explain plan for the insert statement.
http://us.f1.yahoofs.com/users/30a3c...M2j_8AwetM899C
Is there any possibility to reduce the cost for this.
- A unique key exists on the table.
- Table is analyzed with estimate stats for 20 percent.
- Total Records right now 45M +.
Thanx in Advance.
Jr.
I don't know if these suggestions will actually reduce the explain plan's cost #, but ..
Are there other indexes on the table? Key writing will slow things down a bit. I wonder if the cost # goes down if there are fewer number of indexes to build on insert?
Add more CPU's and increase parallelism .. not sure if that will actually reduce the cost #, but it will insert faster .. i/o permitting.
Thanx Ken,
here is a small test run output ...
http://us.f1.yahoofs.com/users/30a3c...Pek_8ASzDCm4Yz
- Could Reduce the Cost with NO PX.
- But Still Cost is about 2000+
- There is only one index on the Table and thats the Unique Key index.
- Not sure the affect of PX when trying to insert 1 Record in context with Time.
Thanx
Jr.
How to load huge data in a table:
Increase freelist for the table. Create necessary extents for the table on different datafiles.
Set the table into nologging mode.
Turn off ARCHIVE LOG.
Drop or disable the index before a load.
After the insert, recreate indexes.
- The table is already set in NOLOGGING modeQuote:
Originally posted by tamilselvan
How to load huge data in a table:
Increase freelist for the table. Create necessary extents for the table on different datafiles.
Set the table into nologging mode.
Turn off ARCHIVE LOG.
Drop or disable the index before a load.
After the insert, recreate indexes.
- Freelists and Initrans are enough set.
- System is in NOARCHIVE LOG Mode
- Enough big Preallocated Extents are available
- And inserts are mandatory dependent on the Unique constraint hence cannot drop and recreate.
So, are there any corners i should look into.
Thanx
Jr.
Then don't concern it.