Re: Slooow Import

From: Harri Kaukovuo <hkaukovu_at_us.oracle.com>
Date: 1996/04/23
Message-ID: <4lh8n1$2u2_at_inet-nntp-gw-1.us.oracle.com>#1/1


jmjm_at_hogpf.ho.att.com (-J.MIKLEWICZ) wrote:

>> ...
>>
>>According the Oracle hotline support import should work in the following way:
>>First the data table is imported and then the index is added. ?????
>>I do not known if I belive this, because import is real slow.
>>
>>Bjarne Ebsen
>>be_at_cci.dk
>>
 

>I've treid to import a large table with indexes = n, i.e. I tell import
>not to import the indexes, and this isn't much faster than a "normal"
>import. I believe the reason is because import works as described above,
>i.e. it imports the data and then creates the indexes. When I import, I
>get a message like "Imported 222,000 rows" which I believe means the
>data import has completed and then import sits there for a while before
>actually completing, which I believe is the index building phase.
>--
 

>Joe Miklewicz 908-949-3061 jmjm_at_hogpa.att.com

This is exactly how it is done; first plain import then index creation.

For faster import you could try these things: - make sure table does not have indexes before you start the import. If does, drop indexes, import the data and either let import to recreate indexes or create yourself after import.

- set COMMIT=Y
- increase BUFFER (try 1024000 for example) size
- place import files on different physical disk than the datafiles are
- if index creation is the problem, you could try increasing
SORT_AREA_SIZE in init.ora.
- import directly from disk if possible (vs. if you have exported to
tape)

By default COMMIT=N which means import will consume a lot of rollback size. Setting COMMIT=Y will commit the rows after each BUFFER size.

  • Harri

hkaukovu_at_us.oracle.com
** The statements and opinions expressed here are my own and ** do not necessarily represent those of Oracle Corporation. Received on Tue Apr 23 1996 - 00:00:00 CEST

Original text of this message