After the loads complete, you can re-create the
table??™s indexes.
In serial Direct Path Loading (PARALLEL=FALSE), SQL*Loader loads data into extents in the
table. If the load process fails before the load completes, some data could be committed to the
table prior to the process failure. In Parallel Data Loading, each load process creates temporary
segments for loading the data. The temporary segments are later merged with the table. If a
Parallel Data Loading process fails before the load completes, the temporary segments will not
have been merged with the table. If the temporary segments have not been merged with the table
being loaded, no data from the load will have been committed to the table.
You can use the SQL*Loader FILE parameter to direct each data loading session to a different
datafile. By directing each loading session to its own datafile, you can balance the I/O load of the
loading processes. Data loading is very I/O intensive and must be distributed across multiple disks
for parallel loading to achieve significant performance improvements over serial loading.
After a Parallel Data Load, each session may attempt to reenable the table??™s constraints. As
long as at least one load session is still underway, attempting to reenable the constraints will fail.
The final loading session to complete should attempt to reenable the constraints, and should
succeed. You should check the status of your constraints after the load completes.
Pages:
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447