Import extremely slow from 9i to 10G

Question

I am preparing a 10.2.0.4 database for testing, (upgrading from 9i) because we are also moving to a new server, I am going to use the exp/imp method. I created a new database after installing and patching the software, then created the necessary tablespaces and users before using imp on the last export taken from the 9i database. Also turned off archive logging for the sake of importing an testing.
This is usually not the fastest process, the dump file is 13gb from the 9i DB, but when I refresh 9i test or dev databases quite often using the same method and do not experience the slowness I am seeing right now in 10G. It has been running for almost 8 hours on one table, which has just over 4 million rows in it, no blob, lob etc.. columns, and nothing of note in the alert log.
9i database is on Windows 2000 advanced server, 10G is on windows 2003 server with plenty of horsepower and RAM, but the cpu is never over 1-3% used
Any ideas here?

Answer 1

Hi,
Try this:
1- Analyze only after import (Analyze=n)
2 -Have de dump to import on a different disk than the DB files.
3- Use commit=n
4- Use only a big rollback segment and put the others offline during the import
5- USe indexes=n and create them after import had finish
6- Use a big buffer during import, example:
(import option - time sec - Improvement %)
commit=y - 120 - 0
commit=y,buffer=64000 - 100 - 17%
commit=n,buffer=30720 - 72 - 40%
commit=n,buffer=64000 - 67 - 44%
7- Pre-create tables before with scripts and use ignore=y
8- use bigger db_cache_size and log_buffer parameters
9- stop archive generation, (ALTER DATABASE NOARCHIVELOG;)
After import is finish, don't forget create all indexes and generate new statistics.
Regards,
Francisco Munoz Alvarez.