Originally posted by pando
To speed up IMPORT

- Use COMMIT=N. This will cause import to commit after each object (table), not after each buffer. This is why one large rollback segment is needed.


from http://metalink.oracle.com/metalink/...T&p_id=93763.1
It is not correct..
During a large table import ..

Exporting from one session:=
14:54:37 SQL> host exp80 XXXXxx/asdf@asdf file=d:\ordrdetl.dmp direct=y constraints=n tables=ordrdetl

14:55:54 SQL> select count(*) from ordrdetl;

COUNT(*)
---------
874350

real: 2610
Importing the same into another user.
14:56:14 SQL> host imp80 backup/xxx@test file=d:\ordrdetl.dmp commit=y full=y

From ANother session of the backup user
I parallely run the count.. It increase ..gradually depending on the Buffer value we have
given

14:57:03 SQL> select count(*) from ordrdetl;

COUNT(*)
---------
33250

real: 16

14:58:13 SQL> /

COUNT(*)
---------
90972

real: 31
14:58:16 SQL> /

COUNT(*)
---------
115178

real: 31
14:58:21 SQL> /

COUNT(*)
---------
147630

real: 32
14:58:27 SQL> /

COUNT(*)
---------
165319

real: 47
14:58:30 SQL> /

COUNT(*)
---------
181146

real: 31
14:58:31 SQL> /

COUNT(*)
---------
182875

real: 63
14:58:33 SQL> show user
USER is "BACKUP"

But during commit=n you cannot even decribe the table..in the other user