CTAS - DOP

Question

Version :
SQL> select * from gv$version;
   INST_ID BANNER
         2 Oracle Database 10g Enterprise Edition Release 10.2.0.4.0 - 64bi
         2 PL/SQL Release 10.2.0.4.0 - Production
         2 CORE 10.2.0.4.0      Production
         2 TNS for IBM/AIX RISC System/6000: Version 10.2.0.4.0 - Productio
         2 NLSRTL Version 10.2.0.4.0 - Production
         1 Oracle Database 10g Enterprise Edition Release 10.2.0.4.0 - 64bi
         1 PL/SQL Release 10.2.0.4.0 - Production
         1 CORE 10.2.0.4.0      Production
         1 TNS for IBM/AIX RISC System/6000: Version 10.2.0.4.0 - Productio
         1 NLSRTL Version 10.2.0.4.0 - Production
10 rows selected.hi experts
I am planning to do CTAS in parallel mode on my production environment which has below configuration.
SQL> show parameter parallel
NAME                                 TYPE        VALUE
fast_start_parallel_rollback         string      LOW
parallel_adaptive_multi_user         boolean     TRUE
parallel_automatic_tuning            boolean     FALSE
parallel_execution_message_size      integer     2152
parallel_instance_group              string
parallel_max_servers                 integer     485
parallel_min_percent                 integer     0
parallel_min_servers                 integer     0
parallel_server                      boolean     TRUE
parallel_server_instances            integer     2
parallel_threads_per_cpu             integer     2
NAME                                 TYPE        VALUE
recovery_parallelism                 integer     0
SQL> show parameter cpu
NAME                                 TYPE        VALUE
cpu_count                            integer     32
parallel_threads_per_cpu             integer     2I have opted for DOP 8 and below is my CTAS
ALTER SESSION FORCE PARALLEL QUERY PARALLEL 8;
ALTER SESSION FORCE PARALLEL DDL PARALLEL 8;
create table r_dummy PARALLEL
as
WITH fq AS
select  tdocstorage_pk
from    fnet_qdocstorage fq
     where fq.batch_id is null
select    bb.docstorageid,
     bb.filenetdocid,
     bb.filename,
     nvl(bb.obsoleteflag, 0) Obsoleteflag,    
     cc.doccategorycode,
     cc.docimagecode DOCFILETYPECODE,
     ee.packetidtypecode,                       
     nvl(ee.docdeliveryflag, 0)  docdeliveryflag
from    closr.t_docstorage bb,
        closr.t_docstoragetype cc,
        closr.t_docstorageattribute ee
where   exists
                select  1
                from    fq
                where   fq.tdocstorage_pk = bb.docstorageid
        and    bb.filenetdocid is not null
        and    bb.docstoragetypecode = cc.docstoragetypecode
        and    bb.docstorageid   = ee.docstorageid (+)
Explain plan :
Plan
1 Every row in the table CLOSR.T_DOCSTORAGETYPE  is read.
2 PX BLOCK ITERATOR
3 PX SEND BROADCAST
4 PX RECEIVE
5 Every row in the table Partitions accessed #1 is read.
6 PX BLOCK ITERATOR
7 PX SEND HASH
8 PX RECEIVE
9 The rows from step 8 were sorted to eliminate duplicate rows.
10 PX JOIN FILTER CREATE
11 Every row in the table CLOSR.T_DOCSTORAGE  is read.
12 PX BLOCK ITERATOR
13 PX JOIN FILTER USE
14 PX SEND HASH
15 PX RECEIVE
16 The result sets from steps 10, 15 were joined (hash).
17 The result sets from steps 4, 16 were joined (hash).
18 PX SEND HASH
19 PX RECEIVE
20 Every row in the table CLOSR.T_DOCSTORAGEATTRIBUTE  is read.
21 PX BLOCK ITERATOR
22 PX SEND HASH
23 PX RECEIVE
24 Rows from step 19 which matched rows from step 23 were returned (hash join).
25 The rows from step 24 were inserted into HLODS.R_DUMMY  using direct-load insert.
26 PX SEND QC (RANDOM)
27 PX COORDINATOR
28 CREATE TABLE STATEMENT Fortunately I've select on all the tables, when I fire this query with DOP as 8 it's taking 2mins to retrive 16653835 rows.(which is good, I guess)
I've few questions here:
1) What would be the optimal DOP for DDL to run when the query is running at a DOP 8
2) Can that be also 8?
3) Is there any way to estimate the time that will be taken by CREATE TABLE if I go with DOP 8?
3) Can I use NOLOGGING here? Does that show any significant difference in time(as we could skip REDO) ?
Our database is in noarchive log mode and the table we are creating here is a just a dummy table, which would be dropped once after we export the dump.
Please help me out, unfortunately I'm not able to test this on other environments because we don't have same config as of PROD.

Answer 1

>
3) Can I use NOLOGGING here? Does that show any significant difference in time(as we could skip REDO) ?
>
Yes - no need to log since you have the source data available if you need it.