Gunnar,
 
It depends on where you are spending most of the wait time.
db file scattered read , for example, can be reduced by setting higher db_file_multiblock_read_count.
Use event 10046 to collect  wait profile.
 
HTH
Vadim
 
-----Original Message-----
From: Gunnar Berglund [mailto:[EMAIL PROTECTED]
Sent: Monday, September 08, 2003 8:49 AM
To: Multiple recipients of list ORACLE-L
Subject: How to speed up index creation

Hi all,
 
we have a couple of huge tables and we  have some performance problems while creating indexes. Here are some methods we already have tried, is there something else:
 

Here's the list of features we already know:

- Using NOLOGGING mode

- Using PARALLEL option => PARALLEL 4 (the server has 4 CPU's)

- Using NOSORT option => not applicable because the rows are not stored in the right order

- Creating an index first as unusable and then rebuilding it => this was tested after the QA cutover but the duration remained unchanged

- Increasing sort_area_size => has not yet been tested

Something else, any help will be appreciated...

 

rgds

g


Want to chat instantly with your online friends? Get the FREE Yahoo! Messenger

Reply via email to