Re: Data Pump and compression

  • From: "Greg Rahn" <greg@xxxxxxxxxxxxxxxxxx>
  • To: piontekdd@xxxxxxxxx
  • Date: Fri, 28 Sep 2007 09:19:03 -0700

Using a pipe w/ gzip seems contradictory to the documentation:

Can I use gzip with Data Pump?
Because Data Pump uses parallel operations to achieve its high
performance, you cannot pipe the output of Data Pump export through
gzip. Starting in Oracle Database 11g, the COMPRESSION parameter can
be used to compress a Data Pump dump file as it is being created. The
COMPRESSION parameter is available as part of the Advanced Compression
Option for Oracle Database 11g.
[http://www.oracle.com/technology/products/database/utilities/pdf/datapump11g2007_quickstart.pdf]

How do I pipe a Data Pump job through gzip?
This compression technique cannot be used with Data Pump, because Data
Pump cannot support named pipes In Oracle Database 10g Release 2, Data
Pump compresses metadata by default Stay tuned for a data compression
solution in a future release.
[http://www.tcoug.org/Archive/Winter2007/TCOUG_Data_Pump.ppt]

Metadata can be quite large.  If you have a range/hash partitioned
table with local indexes and a large number of partitions is can be
quite significant.


On 9/28/07, Bradd Piontek <piontekdd@xxxxxxxxx> wrote:
>
> If you've found a way to do this with datapump, I'd love to see it. (10gR2
> that is).
> Datapump requires directory objects to be created, and is run as a job on
> the database. It doesn't have the same shell interaction like 'exp' and
> 'imp'
>
>
>
> ----- Original Message ----
> From: "Kerber, Andrew W." <Andrew.Kerber@xxxxxxx>
> To: rjoralist@xxxxxxxxxxxxxxxxxxxxx; oracle-l@xxxxxxxxxxxxx
> Sent: Friday, September 28, 2007 10:19:46 AM
> Subject: RE: Data Pump and compression
>
> Shrinking the metadata wouldn't help a lot, its just not that big.
>
> We export through a pipe and use gzip to compress it, and get pretty
> good compression of standard export files,  I expect it would work with
> datapump files also if you are only using one output file.


-- 
Regards,

Greg Rahn
http://structureddata.org
--
//www.freelists.org/webpage/oracle-l


Other related posts: