Given enough CPU and I/O resources, this can significantly reduce the overall loading
times.
Chapter 8: Database Tuning 263
To use Parallel Data Loading, start multiple SQL*Loader sessions using the parallel keyword
(otherwise, SQL*Loader puts an exclusive lock on the table). Each session is an independent
session requiring its own control file. The following listing shows three separate Direct Path loads,
all using the PARALLEL=TRUE parameter on the command line:
sqlload USERID=ME/PASS CONTROL=PART1.CTL DIRECT=TRUE PARALLEL=TRUE
sqlload USERID=ME/PASS CONTROL=PART2.CTL DIRECT=TRUE PARALLEL=TRUE
sqlload USERID=ME/PASS CONTROL=PART3.CTL DIRECT=TRUE PARALLEL=TRUE
Each session creates its own log, bad, and discard files (part1.log, part2.log, part3.log, part1.
bad, part2.bad, and so on) by default. Because you have multiple sessions loading data into the
same table, only the APPEND option is allowed for Parallel Data Loading. The SQL*Loader
REPLACE, TRUNCATE, and INSERT options are not allowed for Parallel Data Loading. If you need
to delete the table??™s data before starting the load, you must manually delete the data (via delete or
truncate commands). You cannot use SQL*Loader to delete the rows automatically if you are
using Parallel Data Loading.
NOTE
If you use Parallel Data Loading, indexes are not maintained by the
SQL*Loader session. Before starting the loading process, you must
drop all indexes on the table and disable all its PRIMARY KEY and
UNIQUE constraints.
Pages:
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446