Re: Bulk loading partitioned tables slower than heap tables?

  • From: amonte <ax.mount@xxxxxxxxx>
  • To: ryan_gaffuri@xxxxxxxxxxx
  • Date: Sat, 24 Jun 2006 00:32:25 +0200

Hi Ryan

The table is hash partitioned into 16 partitions. The tablespace has 16
datafiles. The degree used is 16.

The problems seems with Parallel DML, when I stopped using Parallel DML the
HW space management lock which I had contention with disappeared.

Does anyone know how HW enqueue works?


On 6/23/06, ryan_gaffuri@xxxxxxxxxxx <ryan_gaffuri@xxxxxxxxxxx> wrote: 


What kind of partitioning did you use? How did you spread out the
partitions. did you have them in sepearte datafiles?

In the past I have had problems with full tablescanning hash partitioned
tables if I did not use parallel slaves. It was slower than full scanning a
heap table.

-------------- Original message --------------
From: amonte <ax.mount@xxxxxxxxx>
Hi

I was wondering if anyone has experience bulk loading data to partitioned
tables? I have run some tests and running bulk load (insert append) into
partitioned tables is actually 40% more costy. For example to load up a 80
million rows table it takes around 8 minutes whereas with plain heap table
it only takes 5

Test used:
LMT with 16MB uniform size extent
No ASSM
Parallel DML
Parallel Query
Degree 16

Regards

Alex


Other related posts:

  1. Home
  2. oracle-l
  3. Archive
  4. 06-2006