Speeding up the expdp job with PARALLEL and FILESIZE parameters

Question

Daily we are going to back up 6 schemas with a total size of 80 GB.
From oracle documentation I gather that PARALLEL servers work well when we split the dump file because each slave process can work with a separate file.
But I am not sure how many parallel processes should be spawned and the how many files this dump file has to be split?
The expdp command we are planning to use
expdp userid=\'/ as sysdba\' SCHEMAS = schema1,schema2,schema3,schema4,schema5,schema6  DUMPFILE=composite_schemas_expdp.dmp LOGFILE=composite_schemas_expdp.log  DIRECTORY=dpump_dir2 PARALLEL=3Related info:
11.2.0.2
Solaris 10 (x86_64) running on HP Proliant Machine
8 CPU with 32gb RAM
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_degree_limit                string      CPU
parallel_degree_policy               string      MANUAL
parallel_execution_message_size      integer     16384
parallel_force_local                 boolean     TRUE
parallel_instance_group              string
parallel_io_cap_enabled              boolean     FALSE
parallel_max_servers                 integer     32
parallel_min_percent                 integer     0
NAME                                 TYPE        VALUE
parallel_min_servers                 integer     0
parallel_min_time_threshold          string      AUTO
parallel_server                      boolean     TRUE
parallel_server_instances            integer     2
parallel_servers_target              integer     32
parallel_threads_per_cpu             integer     2
recovery_parallelism                 integer     0

Answer 1

I just did a test export of a 16gb schema (schema size confirmed from dba_segments).
I didn't specify the FILESIZE parameter. Instead I specified PARALLEL = 5 with DUMPFILE=tstschema_expdp%U.dmp hoping that expdp will create 5 different dumpfiles for the 5 slave processes to work with. But expdp created only 4 files.
Export: Release 11.2.0.1.0 - Production on Mon Aug 6 10:30:57 2012
Copyright (c) 1982, 2009, Oracle and/or its affiliates.  All rights reserved.
Connected to: Oracle Database 11g Enterprise Edition Release 11.2.0.1.0 - 64bit Production
With the Partitioning, OLAP, Data Mining and Real Application Testing options
Starting "EXPIMP_USER"."SYS_EXPORT_SCHEMA_01":  expimp_user/******** SCHEMAS = TSTSCHEMA DUMPFILE=tstschema_expdp%U.dmp LOGFILE=test_expdp.log DIRECTORY=dpump_dir PARALLEL=5
Estimate in progress using BLOCKS method...
Processing object type SCHEMA_EXPORT/TABLE/TABLE_DATA
Total estimation using BLOCKS method: 9.089 GB
Processing object type SCHEMA_EXPORT/USER
Processing object type SCHEMA_EXPORT/SYSTEM_GRANT
Processing object type SCHEMA_EXPORT/ROLE_GRANT
Processing object type SCHEMA_EXPORT/DEFAULT_ROLE
Processing object type SCHEMA_EXPORT/PRE_SCHEMA/PROCACT_SCHEMA
. . exported "TSTSCHEMA"."FTB_RES_RCS"                      14.75 MB   13860 rows
Processing object type SCHEMA_EXPORT/SYNONYM/SYNONYM
Processing object type SCHEMA_EXPORT/TABLE/TABLE
Processing object type SCHEMA_EXPORT/TABLE/GRANT/OWNER_GRANT/OBJECT_GRANT
Processing object type SCHEMA_EXPORT/TABLE/INDEX/INDEX
Processing object type SCHEMA_EXPORT/TABLE/CONSTRAINT/CONSTRAINT
Processing object type SCHEMA_EXPORT/TABLE/INDEX/STATISTICS/INDEX_STATISTICS
Processing object type SCHEMA_EXPORT/TABLE/INDEX/FUNCTIONAL_AND_BITMAP/INDEX
Processing object type SCHEMA_EXPORT/TABLE/INDEX/STATISTICS/FUNCTIONAL_AND_BITMAP/INDEX_STATISTICS
Processing object type SCHEMA_EXPORT/TABLE/STATISTICS/TABLE_STATISTICS
. . exported "TSTSCHEMA"."CASR_HAP_DTL"                      824.9 MB 1613143 rows
. . exported "TSTSCHEMA"."JOHN"                                  0 KB       0 rows
. . exported "TSTSCHEMA"."DCVERSO"                           7.210 GB 7502854 rows
Master table "EXPIMP_USER"."SYS_EXPORT_SCHEMA_01" successfully loaded/unloaded
Dump file set for EXPIMP_USER.SYS_EXPORT_SCHEMA_01 is:
  /u05/bkp_DUMP/tstschema_expdp01.dmp
  /u05/bkp_DUMP/tstschema_expdp02.dmp
  /u05/bkp_DUMP/tstschema_expdp03.dmp
  /u05/bkp_DUMP/tstschema_expdp04.dmp
Job "EXPIMP_USER"."SYS_EXPORT_SCHEMA_01" successfully completed at 11:07:36My conclusion:
If you want to keep Number of dumpfiles to be generated and Number of Parallel processes to be N (same) then use FILESIZE
ie.
Say you want to export an 80gb schema and if you setting PARALLEL = 5 , it is better to specify FILESIZE after doing the arithmetic.
Divide 80/5 = 16 . So put the FILESIZE = 16g so that expdp will create 5 different files to 5 different 5 slave processes to work with.