I was not able to see an appreciable gains with commit=y, just curious why the docs would say that and if others had different experiences with seeing gains from commit=y.
I little more background. We are moving our data center and in the process going from windooze to unix. In our go-live weekend every minute counts if you consider backup to tape, ship tape, retore etc...
Anyhow, our db dump is 30gb so what I did was split up the export in to several parts. We have one app user (PeopleSoft) who is sysadm so I did the following. Once I produced multiple export files and ran 6 import sessions in parallel my import went from 4 hours to 45 minutes.
CREATE OR REPLACE DIRECTORY export_dir AS 'c:\temp';
CREATE OR REPLACE PROCEDURE export_for_refresh
v_counter NUMBER :=0;
v_logfile_number NUMBER :=1 ;
v_output_file1 := utl_file.fopen('EXPORT_DIR', 'export_for_refresh_file_' || v_logfile_number || '.txt', 'W');
FOR cursor_table IN (SELECT table_name FROM all_tables WHERE owner = 'SYSADM')
utl_file.put_line (v_output_file1, cursor_table.table_name);
IF v_counter =1000 THEN
v_output_file1:= utl_file.fopen('EXPORT_DIR', 'export_for_refresh_file_' || v_logfile_number || '.txt', 'W');
v_counter := v_counter+1;