Re: 11gR2 Data Pump question - parallel setting
- From: rjamya <rjamya@xxxxxxxxx>
- To: JC1706@xxxxxxx
- Date: Fri, 27 Jul 2012 07:21:03 -0400
Jon,
index/constraints etc will run parallel only if they had parallel
setting at the time of expdp. parallelism is mainly for
creating/loading tables, rest of metadata is always handled by a
single thread and so is index/constraint creation etc.In fact you can
split index creation scripts by table and them submit them in parallel
via different sessions and then do same for constraints if applicable.
If you want to do real parallelism, do what Greg suggested, create
script and then tweak it to create indexes in parallel (and then alter
to reset parallelism if you need to). Based on requirements, you may
be able to speed up constraint creation by using novalidate (ymmv).
Broadly speaking the parallelism doesn't affect
index/constraint/metadata part (not sure of 11203 or 12c yet). Also
for loading data I am not really sure if having parallel=x where x >>
number of your dumpfiles will benefit greatly. However x <
number_of_datafiles will slow things down a bit in my experience.
Raj
On Thu, Jul 26, 2012 at 6:14 PM, CRISLER, JON A <JC1706@xxxxxxx> wrote:
> Is there any benefit to setting a Data Pump import to parallel=12 if the
> export was run with parallel=8 ? It seems to accept higher values, and
> spawns 12 workers. My goal is to get the items like index creation,
> constraints etc. to run faster. The source system has 8 cpu's, the target
> system 24 cpus.
--
//www.freelists.org/webpage/oracle-l
Other related posts:
