|
See Also
When import task performance
is poor, this is usually because routines are repeated over and
over for each import record. You may be able to improve import
performance by separately executing routines that are not
record-bound, and by removing routines that are not absolutely
necessary.
These routines include:
|
·
|
Computing values such as totals, which are re-adjusted for each
import record but could also be calculated in one go AFTER
import. |
|
·
|
Row-bound processing that can be performed IN the database at a
later time. |
|
·
|
Creation of unnecessary status files. |
|
·
|
SQL processing that requires missing indexes. |
Furthermore, you can
deactivate certain constraints and relational checks.
To improve import task
performance:
| 1.
|
Consider computing values as a separate routine outside the
import task, if possible. |
Perform the calculation as a
separate routine, using a SQL task or corrective constraint.
| 2.
|
Consider delaying part of the import task until later. Perform
a straight import, marking records as needing to be
post-processed. |
Perform post-processing as a
separate routine on the marked records, using a SQL task or
corrective constraint.
| 3.
|
If you do not require import status files, disable their
creation. |
| 4.
|
Do NOT attempt to cut down on database SQL processing as this
is usually not where the performance bottleneck is. |
|
·
|
However, DO consider if any of these statements are slow
because they require indexes that were not created. If so, create
additional indexes, for instance, on much-used foreign key
values. |
|
