Oracle® Retail Predictive Application Server Cloud Edition Administration Guide Release 23.1.201.0 F80305-03 |
|
![]() Previous |
![]() Next |
This chapter describes the best practices for PDS.
This section describes measure to fact mapping and optimal fact grouping.
When working with multiple applications, a configurator is expected to carefully map the application measures to facts, especially for the measures/facts shared by multiple applications. Incorrect or missing mapping may not only unnecessarily increase the number of facts, but also result in issues in data integration later at run time. For example, app1/measureA and app2/measureB are meant to be a shared measure and must be mapped to the same fact. If this mapping is not identified, instead of having a shared fact1, the configurator incorrectly creates two facts. Such incorrect mapping not only unnecessarily creates more facts than needed, but also breaks the integration between app1/measureA and app2/measureB.
In PDS, facts are grouped and facts in the same group are stored in the same fact table. Optimal fact grouping is automatically done by ConfigTools based on the facts' base intersection, batch and workbook commit rules, facts' input source, and so on. (see Oracle Retail Predictive Application Server Cloud Edition Configuration Tools User Guide for details).
Fact grouping has a direct performance impact on data loading, batch calculation, and workbook commit. The configurator is expected to carefully configure all rules and rule groups, assign measures' base intersections, and file name attributes, which defines the measures' input source.
Facts can be loaded from flat files or RDX interfaces. To achieve optimal performance, each input file or RDX interface must contain all measures/facts with the same file name attribute. In other words, each file or RDX interface must cover all facts of the target fact tables. (Mapping of one file/interface to two+ fact tables is fine.)
Free expressions are supported in batch control file, but their execution is not optimal, for two reasons. They are translated into PL/SQL blocks and executed expression by expression, in the order that they appear in the batch control file. The free expressions are not taken into consideration in optimal fact grouping as they do not appear as batch rule groups in the application configuration.
It is recommended that the users minimize the use of free expressions. An alternative way is to configure them as rule group(s).