| Oracle® Fusion Middleware Administrator's Guide for Oracle Business Intelligence Applications 11g Release 1 (11.1.1.7) Part Number E37988-01 |
|
|
PDF · Mobi · ePub |
| Oracle® Fusion Middleware Administrator's Guide for Oracle Business Intelligence Applications 11g Release 1 (11.1.1.7) Part Number E37988-01 |
|
|
PDF · Mobi · ePub |
This chapter describes concepts and techniques for customizing the ETL functionality in Oracle Business Intelligence Applications.
This chapter contains the following topics:
Section 1.1, "Overview of Customization in Oracle Business Intelligence Applications"
Section 1.2, "Category 1 Customizations: Adding Columns to Existing Fact or Dimension Tables"
Section 1.3, "Category 2 Customizations: Adding Additional Tables"
Section 1.5, "Customizing Stored Lookups and Adding Indexes"
This section provides an overview of customization in Oracle Business Intelligence Applications, and contains the following topics:
Section 1.1.1, "What is Customization in Oracle Business Intelligence Applications?"
Section 1.1.3, "About the Impact of Patch Installation on Customizations"
In Oracle Business Intelligence Applications, customization is defined as changing the preconfigured behavior to enable you to analyze new information in your business intelligence dashboards. For example, you might want to add a column to a dashboard by extracting data from the field HZ_CUST_ACCOUNTS.ATTRIBUTE1 and storing it in the Oracle Business Analytics Warehouse in the X_ACCOUNT_LOG field.
The type of data source that you have determines the type of customization that you can do. Data sources can be one of the following types:
Packaged applications (for example, Oracle EBS), which use prepackaged adapters.
Non-packaged data sources, which use the Universal adapter.
Customizations are grouped into the following categories:
Category 1. In a Category 1 customization, you add additional columns from source systems that have pre-packaged adapters and load the data into existing Oracle Business Analytics Warehouse tables. For more information about performing Category 1 customizations, see Section 1.2, "Category 1 Customizations: Adding Columns to Existing Fact or Dimension Tables".
Category 2. In a Category 2 customization, you use pre-packaged adapters to add new fact or dimension tables to the Oracle Business Analytics Warehouse. Category 2 customizations normally require that you build new SDE and SIL mappings. For more information about performing Category 2 customizations, see Section 1.3, "Category 2 Customizations: Adding Additional Tables".
Category 3. In a Category 3 customization, you use the Universal adapter to load data from sources that do not have pre-packaged adapters. For more information about performing Category 3 customizations, see Section 1.4, "Category 3 Customizations: Adding New Data as a Whole Row into a Standard Dimension Table".
The figure below summarizes the category of customization that you can perform for each type of data source and type of modification.
For detailed information about tables and naming conventions, see Oracle Business Analytics Warehouse Data Model Reference.
When you customize ETL Packages and Interfaces, you usually work in the \Oracle BI Applications\Mappings folder in the Projects view in ODI Studio's Designer Navigator.
Note: The customization methodology is to make a copy of the ETL task and version both the original and copy while a datastore is simply versioned. These versions allow you to revert functionality if required as well as identify changes that have been introduced through customization, patches or upgrades.
This chapter explains how to customize your ETL functionality, after you have performed a Business Analysis and Technical Analysis. This chapter does not cover the other typical tasks that you need to perform, as follows:
Business Analysis - before you start customization, you typically analyze your current BI dashboards to determine the changes you need to support your business or organization.
Technical Analysis - when you have identified your business requirements, you need to determine the technical changes you need to make, by identifying source tables, staging tables, target tables, and ODI Packages and Interfaces that you need to modify.
RPD Modification - having made the customizations in the ETL functionality, you need to modify your RPD to expose the new data in your dashboards. For more information about RPD modification, refer to the Oracle Business Intelligence Enterprise Edition documentation library.
This section explains what you must do to re-apply a customization to an object that has been patched. For example, if you install an Oracle Business Intelligence Applications patch that modifies the Supply Chain and Order Management application, you might need to manually re-apply customizations that you have made to the Supply Chain and Order Management application.
As part of customizing an ETL task (including interfaces and package under a specific task folder), you copy the task folder to be customized, version the original and version the copy. Any patches are applied to the current version of the original task. Leverage ODI's version compare utility to identify the changes introduced by the patch. The copy is also versioned so that any changes introduced can be isolated. Compare any changes with those introduced by the patch and verify there is no conflict, then manually apply the same changes introduced by the patch to the customized ETL tasks. For information about modifying and versioning ETL customizations, refer to Section 1.2.2, "Typical Steps to Extend Mappings in the Oracle Business Analytics Warehouse".
A patch only installs changed repository objects, not the whole Work Repository. Therefore, you only need to re-apply customizations to mappings that have been changed by the patch. For example, if a patch only modifies the Supply Chain and Order Management application, you only need to manually re-apply customizations that you have made to the Supply Chain and Order Management application. Customizations in other applications are not affected by the patch.
All customization steps have you create a 'Custom' adaptor folder where customized ETL tasks are stored. This is not required but is considered a best practice to make identifying customized content easier.
Category 1 customizations add additional columns from source systems that have pre-packaged adapters and load the data into existing Oracle Business Analytics Warehouse tables.
This section contains the following topics:
Section 1.2.2, "Typical Steps to Extend Mappings in the Oracle Business Analytics Warehouse"
Section 1.2.3, "Other Types of Customizations Requiring Special Handling"
Category 1 customizations involve extracting additional columns from source systems for which pre-packaged adapters are included (for example, Oracle eBusiness Suite) and loading the data into existing Oracle Business Analytics Warehouse tables. For Category 1 customizations, data can also come from non-packaged sources, but this section assumes that the sources have already been mapped with a Universal adapter and only need to be extended to capture additional columns. (The initial mapping of a Universal adapter is considered a Category 3 customization. For information, see Section 1.4, "Category 3 Customizations: Adding New Data as a Whole Row into a Standard Dimension Table".)
In order to see additional columns in the Oracle Business Analytics Warehouse, the columns must first be passed through the ETL process. The existing mappings and tables are extensible. Oracle Business Intelligence Applications provides a methodology to extend preconfigured mappings to include these additional columns and load the data into existing tables.
Oracle Business Intelligence Applications recognizes two types of customization: extension and modification. The supported extension logic allows you to add to existing objects. For example, you can extract additional columns from a source, pass them through existing mappings, and populate new columns added to an existing table. Generally, Oracle Business Intelligence Applications does not allow you to modify existing logic or columns. You should not change existing calculations to use different columns, and you should not remap existing columns to be loaded from different sources.
For example, if you want to calculate revenue differently from the existing logic, you should create a new column (for example, X_REVENUE) and populate it with a custom mapping expression. You can then remap the Oracle Business Intelligence repository to point to the new X_REVENUE column.
Most datastores have a single placeholder column named X_CUSTOM. Each ETL task has mapping expressions to populate this column. These serve as templates for customizing ODI datastores and interfaces. When creating new custom columns, follow the naming convention of including the X_ prefix to help distinguish custom columns.
In the figure below, the preconfigured logic is shaded in gray. You should not modify anything contained within these objects. You should add customizations to existing objects rather than creating new packages and interfaces, which allows them to run parallel to the existing logic.
The most common reason for extending the Oracle Business Analytics Warehouse is to extract existing columns from a source system and map them to an existing Oracle Business Analytics Warehouse table (either fact or dimension). This type of change typically requires you to extend the interfaces within a SIL package. If the data is coming from a packaged source, then you will also need to extend the interfaces within an appropriate SDE adapter package. If the data is coming from a non-packaged source, then you must use a Universal adapter package. If an appropriate package does not already exist, you will need to create a Universal adapter package with interfaces.
To extend an ODI package in the Oracle Business Analytics Warehouse:
Create new SDE and SIL Adaptor folders (do not copy existing Adaptor folder as this will copy all subfolders). Rename folders to include 'Custom' or some other useful identifier in the name, and set Release Tag to match that of the existing Adaptor folder. Do this for both the SDE and SIL folders.
Right-click the Mappings folder and select New Sub-Folder.
Set Name as CUSTOM _<Original Folder Name>. For example, CUSTOM_SDE_ORA11510_Adaptor, CUSTOM_SILOS represent custom SDE and SIL folders.
Click the Connect Navigator button in the Designer tab.
Select Edit Release Tags.
Select the release tag that corresponds to your source. For example, EBS_11_5_10.
Select the custom SDE folder you created and add it to the release tag.
Click Next.
Click Finish.
Repeat the above steps for the CUSTOM_SILOS folder, associating it with the BIA_11 Release Tag.
Enable versioning for the preconfigured Task Folder to be customized. The version comment should indicate this is the base version of the task. Subsequent patches applied to this task in the future would require increasing the version in the comment so that it can be compared to the original task to identify any changes.
Right-click the Task folder and select Version > Create Version.
Accept the default version number, 1.0.0.0.
Add a description indicating that this is the original version of this task.
Duplicate the Task folder to be customized by copying it. Cut and paste the copied task folder to the Custom adaptor, and rename it to remove the 'Copy of…' prefix.
Using the same method as in step 2, enable versioning of copied Task folder. The version comment should indicate this is the original version. This versioning enables comparison of the customized task to a copy of the original version to determine all changes that have been introduced.
Create another version of the copied task. The version comment should indicate this is the customized version. Use the same steps as above.
Version the Model that the datastore to be customized exists in, for example, Oracle BI Applications. Submodels and datastores cannot be versioned. The version comment should indicate this is the base or original version.
Create a new version of the model, with a version comment indicating that this is where customizations are introduced. The models can now be compared to show differences. If the model ever needs to be patched, the model should be versioned again so that the patched version can be compared to the custom and original version.
Apply customizations to the datastore and task. Customizations should be additive as much as possible rather than overwriting existing content. For example, if you don't like the way a particular column is calculated, add a new custom column and map it in the way you prefer. In the RPD, have the logical column point to this new custom column rather than the original column.
Prior to generating scenarios, ensure the 'Scenario Naming Convention' User Parameter has a value of %FOLDER_NAME(2)%_%OBJECT_NAME%
Generate scenarios for any new Adaptors, using the option to generate the scenario as if all underlying objects are materialized. The scenario will be generated reflecting the custom adaptor name. In the future if you make changes to any of the interfaces or the package, you can either regenerate the existing scenario or generate a new scenario. Unless you need separate scenarios, it is recommended that you regenerate the existing scenario. To do this, right-click the scenario and select Regenerate.
Generate the Load Plan.
Update Load Plan steps in the generated Load Plan to reference the custom scenario.
Execute the Load Plan.
This section contains the following topics:
The BI Applications ETL process supports Type I and Type II slowly changing dimension behavior. Some dimensions are enabled only for Type I behavior while other dimensions are enabled to also support Type II behavior. Of those dimensions that support Type II behavior, different dimension attributes have different Slowly Changing behavior including some attributes being treated as Type I.To enable or disable Type II behavior associated with a dimension:
Note: Modifying the Type-II tracking logic is the only change that you should make to shipped logic.
To modify a Category 2 SCD Trigger:
In ODI Designer, modify the dimension datastore.
In the Models view, expand the 'Oracle BI Applications' folder, Oracle BI Applications (Model), and Dimension (Submodel).
Double-click the Dimension table.
In the Definition tab, change the OLAP type value to either Dimension (only supports Type I changes) or Slowly Changing Dimension (supports Type II changes).
Modify the SIL Dimension Task.
Navigate to the SIL task that populates this dimension.
Double-click the 'Main' interface.
In the Flow subtab, select the 'Target (ORACLE_BI_APPLICATIONS) window.
If the Property Window is not visible, open it by clicking the Menu Options View – Property Inspector.
Change the IKM Selector value to 'IKM BIAPPS Oracle Slowly Changing Dimension' if enabling Type II behavior or 'IKM BIAPPS Oracle Incremental Update' if removing Type II behavior.
Regenerate the scenario.
The following describes how to modify which columns are treated as Type I or Type II in a dimension that is configured to support Type II behavior. If a dimension is configured to support only Type I behavior, the following changes will have no effect as all columns are treated as Type I.
To enable or disable Type II behavior associated with a dimension:
In ODI Designer, modify the dimension datastore. In the Models view, expand the 'Oracle BI Applications' folder, Oracle BI Applications (Model), Dimension (Submodel), and Columns.
Double-click the column whose SCD behavior you want to change.
In the Description subtab's 'Slowly Changing Dimensions Behavior' drop-down list, select the column behavior. To implement Type I behavior, select Overwrite on Change. To implement Type II behavior, select Add Row on Change.
If enabling Type II behavior for a custom dimension, be sure to set columns as follows:
ROW_WID - Surrogate Key
INTEGRATION_ID, DATASOURCE_NUM_ID - Natural Key
CURRENT_FLG - Current Record Flag
EFFECTIVE_FROM_DT - Starting Timestamp
EFFECTIVE_TO_DT - Ending Timestamp
This section explains how to add a dimension to an existing fact, adding a dimension and dimension staging datastores as well as associated SDE and SIL processes, which also requires extending the fact and fact staging tables to reflect the association with the new dimension. This section includes the following topics:
Create the custom dimension datastores and tasks. Create a WC_<dimension name>_D datastore under the 'Oracle BI Applications – Dimension' model. Create a WC_<dimension name>_DS datastore under the 'Oracle BI Applications – Dimension Stage' model. Use the WC_SAMPLE_DS and WC_SAMPLE_D datastores as templates. These datastores include all required system columns. Custom tables should follow the WC_ naming convention to help distinguish from shipped tables.
Note:
The specific submodel that a table belongs to drives the table maintenance behavior. For example, tables in the 'Dimension Stage' submodel will always be truncated at each ETL run while tables in the 'Dimension' submodel are truncated only during a Full ETL run. Do not create a 'Custom' submodel to place your datastores as table maintenance will not be implemented properly for tables in such a submodel.
As described below, a dimension can be defined either in ODI, generating DDL to create the table in the database, or by defining the table in the database and importing the definition into ODI using the BI Applications RKM. If you use the RKM, the imported table is automatically placed in the 'Other' submodel and needs to be moved into the 'Dimension Staging' and 'Dimension' submodels as appropriate. Also, the OLAP type will need to be set for the dimension to 'Dimension' or 'Slowly Changing Dimension' as appropriate.
Manually Create the Dimension Tables in ODI
To create the dimension and tasks manually using ODI:
In Designer, navigate to Models > Oracle BI Applications (Folder) > Oracle BI Applications (Model) > Dimension Stage (Submodel).
Right-click the WC_SAMPLE_DS datastore and select Duplicate Selection.
Double-click the new datastore and rename it. Name and Resource Name should match the actual table name. Alias can be the same or a more user friendly value.
In the Columns subtab, add all columns.
Repeat the same steps to create the Dimension Table by copying the WC_SAMPLE_D datastore under the Dimensions submodel.
For the dimension table, set the OLAP type to either Dimension if this is a Type I dimension or to Slowly Changing Dimension if this is a Type II dimension.
Import Custom Dimension Tables into ODI
To import custom dimension tables into ODI:
In Designer, navigate to Models > Oracle BI Applications (Folder) and double-click the Oracle BI Applications Model.
In the Reverse Engineer subtab, indicate the tables to be imported under the LIST_OF_TABLES option. To import multiple tables, provide a comma-separated list.
Click the Reverse Engineer button to start a session that imports the table(s) into ODI
The Reverse Engineer process places all tables in the Other submodel. Drag and drop W_%_DS tables into the Dimension Stage submodel and the W_%_D table into the Dimension submodel.
Double-click the new dimension datastore and set the OLAP type to either Dimension if this is a Type I dimension or to Slowly Changing Dimension if this is a Type II dimension.
Create an ODI Sequence for the Custom Dimension
Create an ODI sequence for the custom dimension. A database sequence is used to populate the ROW_WID column of the dimension. The Generate DDL procedure is used to generate the DDL required to create the database trigger in the database. Use WC_SAMPLE_D_SEQ as a template.
In Designer, navigate to Projects > BI Apps Project > Sequences.
Right-click the Sequence folder and select New Sequence.
Set name to <Dimension Name>_SEQ.
Select the Native sequence radio button.
Set the Schema to DW_BIAPPS11G.
Generally, the Native sequence name should match the ODI name unless this causes the name length to exceed 30 characters, in which case, you can shorten the name to meet this limit. This is the name of the database trigger created to populate the ROW_WID column.
Generate the DDL to create the table in the database. Note: If you manually created the dimension in ODI, this will generate the DDL to create both the table and sequence. If you imported the dimension into ODI, this will generate the DDL to create the sequence only.
Create SDE and SIL tasks in the Custom SDE and SIL adaptor folders. Use the SDE_<Product Line Code>_SampleDimension and SIL_SampleDimension tasks as a template. These tasks include the logic required to populate the system columns. Finally, generate scenarios for these tasks.
Add Load Plan step to the '3 SDE Dims X_CUSTOM_DIM <Product Line Version Code>' Load Plan Component.
In Designer, navigate to Load Plans and Scenarios > BIAPPS Load Plan > Load Plan Dev Components > SDE - <Product Line Version Code> and double-click the '3 SDE Dims X_CUSTOM_DIM <Product Line Version Code>' Load Plan Component.
In the Steps subtab, select the 'X_CUSTOM_DIM' step.
Click the green '+' symbol near the top right and select Run Scenario Step.
Provide the Scenario Name, set the Version as -1, and set the Step Name to match the Task name . Set the Restart Type to 'Restart from failed step.'
The Fact related datastores and tasks must be extended to reflect the new dimension. Both the W_<Fact Name>_FS and W_<Fact Name>_F datastores must be extended.
To customize fact datastores and tasks:
Extend the Fact Staging datastore by adding an ID column that follows the naming convention X_<name>_ID and datatype VARCHAR2(80).
The Oracle BI Applications Model should already be versioned.
Navigate to Models > Oracle BI Applications (Folder) > Oracle BI Applications (Model) > Fact Stage (Submodel) and double-click the Fact Staging Table.
In the Columns subtab, select the 'X_CUSTOM' column.
Click the green '+' symbol to add a column below the X_CUSTOM column.
Extend the Fact datastore by adding a WID column that follows the naming convention X_<name>_WID and datatype NUMBER(10). Follow the same steps as above to add a column to the fact datastore.
Add a foreign key constraint to the fact table that refers to the custom dimension table created previously. The foreign key constraint ensures the Custom SIL task is included in the generated load plan. The Custom SDE task is included in the generated load plan because it populates the staging table that is used as a source for the custom SIL task.
Drill into the Fact datastore.
Right-click the Constraints subfolder below the Fact datastore and select New Reference.
The naming convention is FK_<Fact Table>_<Dimension Table>. If there are multiple WID columns that need to reference the same dimension table, enumerate each with a numeric suffix, for example, FK_WC_CUSTOM_F_WC_CUSTOM_D1. Type must be 'User Reference'.
Select the Custom Dimension from the Table drop-down list.
In the Columns subtab, click the green '+' symbol to add a new column.
For the Foreign Table column, select the custom WID column in the fact table. For the Primary Table column, select the ROW_WID column in the dimension table.
Add a non-unique bitmap index on the X_<name>_WID column.
Drill into the Fact datastore.
Right-click the Constraints subfolder below the Fact datastore and select New Key.
The naming convention is <Fact Table>_F<n>. Enumerate each of these indexes with a numeric suffix, for example, WC_CUSTOM_F1.
Select the Not Unique Index radio button.
In the Columns subtab, add the WID column using the shuttle button.
In the Control subtab, check the Defined in the Database and Active check boxes.
In the Flexfields subtab, set the index type value to QUERY and the bitmap index value to Y.
Modify the Fact SDE task. Pass the value from the source table to the custom X_<name>_ID column in the staging table. In the mapping expression, include any necessary conversion functions to get the data into the VARCHAR2(80) format.
Modify the Fact SIL task. Add logic to retrieve the ROW_WID value from the custom dimension. This is usually done in one of the following ways. There is no significant difference between these two methods:
Add the dimension as a source in the SQ temp interface. Join on the fact table's ID column and the dimension table's INTEGRATION_ID column and the fact and dimension DATASOURCE_NUM_ID columns. If the dimension is a Type II dimension, include a range join on the fact's canonical date between the dimension's effective dates. Configure the join as a Left Outer Join. Pass the ROW_WID column as an output.
Add the dimension as a lookup in the main interface. The Lookup join is on the fact table's ID column and the dimension table's INTEGRATION_ID column and the fact and dimension DATASOURCE_NUM_ID columns. If the dimension is a Type II dimension, include a range join on the fact's canonical date between the dimension's effective dates. Configure the Lookup Type as 'SQL left-outer join in the from clause'.
In the mapping expression to populate the custom WID column in the main interface, embed the ROW_WID column from the dimension table in a function that defaults NULL values to 0. For example, COALESCE(SQ_W_AP_HOLDS_FS.PURCHASE_ORG_WID,0)
This use case is similar to adding a regular Dimension to a fact but in this case, a Date dimension is used. There are several Date related dimension, each representing dates in a different manner (fiscal, enterprise, and so on) and different granularities (day, week, month, etc.).
Joins between a fact and Date dimension table are performed on a Date specific WID column. The Date WID column is a 'smart key' value that represents the date in YYYYMMDD format. There is no need to do a lookup to resolve the ROW_WID of the Date dimension, rather you pass the Date column through the ETL process and convert it to this format.
Each fact table has exactly one 'canonical' Date specific WID column. This is the primary date used to drive various date-related calculations. There is no particular metadata to identify this column but lookups to effective dated tables will use this column in the ETL and various date-related expressions in the RPD will also use this column. All packaged fact tables have a single canonical date already identified. When creating custom fact tables, one Date WID column should be nominated as the canonical date and consistently used.
Follow the same steps as adding a dimension to a fact with the following changes. There is no need to create a custom SDE as we use the existing Date dimension.
Customize Fact Datastores and Tasks
The Fact related datastores and tasks must be extended to reflect the new dimensionality. Both the W_<Fact Name>_FS and W_<Fact Name>_F datastores must be extended.
Extend the Fact Staging datastore by adding a DT column that follows the naming convention X_<name>_DT. This column should have the format DATE(7).
Extend the Fact datastore by adding both custom DT and DT_WID columns. These follow the naming convention X_<name>_DT and X_<name>_DT_WID.
Add a foreign key constraint to the Date dimension or dimensions. If there are multiple WID columns that need to reference the same date dimension table, enumerate each with a numeric suffix.
Modify the Fact SDE task. Pass the value from the source table to the custom X_<name>_DT column in the staging table. Apply any conversions required to get the data into DATE format.
Modify the Fact SIL task. Pass the X_<name>_DT value from the staging table to the corresponding column in the fact table. In the mapping expression to populate the custom X_<name>_DT_WID column in the main interface, embed the DT column in a function that calculates the DT_WID value, defaulting to 0 when the supplied DT value is NULL. For example, CALCULATE_DT_WID_DFLT(SQ_W_AP_HOLDS_FS.HOLD_DATE,0)
Category 2 customizations use pre-packaged adapters to add new fact or dimension tables to the Oracle Business Analytics Warehouse.
This section contains the following topics:
This section relates to building entirely new tables that will be loaded with data from a source table that is not already extracted from. For example, you might want to create a new Project dimension table. In this case, you create new dimension and staging tables as well as new extract and load ETL mappings.
When creating a new custom table, use the prefix WC_ to help distinguish custom tables from tables provided by Oracle as well as to avoid naming conflicts in case Oracle later releases a table with a similar name. For example, for your Project dimension you might create a WC_PROJECT_DS and a WC_PROJECT_D table.
When you create a new dimension or fact table, use the required system columns that are part of each of the Oracle Business Analytics Warehouse tables to maintain consistency and enable you to reference existing table structures. When you create a new table, you need to define the table and indices in ODI Designer Models area first. The destination model for the Oracle Business Analytics Warehouse is 'Oracle BI Applications'.
For custom staging tables, the following columns are required:
INTEGRATION_ID. Stores the primary key or the unique identifier of a record as in the source table.
DATASOURCE_NUM_ID. Stores the data source from which the data is extracted.
For dimension and fact tables, the required columns are the INTEGRATION_ID and DATASOURCE_NUM_ID columns as well as the following:
ROW_WID. A sequence number generated during the ETL process, which is used as a unique identifier for the Oracle Business Analytics Warehouse.
ETL_PROC_WID. Stores the ID of the ETL process information.
The tables in the Oracle Business Analytics Warehouse schema have DATASOURCE_NUM_ID as part of their unique user key. While the transactional application normally ensures that a primary key is unique, it is possible that a primary key is duplicated between transactional systems. To avoid problems when loading this data into the data warehouse, uniqueness is ensured by including the DATASOURCE_NUM_ID as part of the user key. This means that the rows can be loaded in the same data warehouse tables from different sources if this column is given a different value for each data source.
This section contains additional miscellaneous information about customization in Oracle Business Intelligence Applications
For loading new fact and dimension tables, design a custom process on the source side to detect the new and modified records. The SDE process should be designed to load only the changed data (new and modified). If the data is loaded without the incremental process, the data that was previously loaded will be erroneously updated again. For example, the logic in the preconfigured SIL mappings looks up the destination tables based on the INTEGRATION_ID and DATASOURCE_NUM_ID and returns the ROW_WID if the combination exists, in which case it updates the record. If the lookup returns null, it inserts the record instead. In some cases, last update date(s) stored in target tables are also compared in addition to the columns specified above to determine insert or update. Look at the similar mappings in the preconfigured folder for more details.
Staging tables typically do not require any indices. Use care to determine if indices are required on staging tables. Create indices on all the columns that the ETL will use for dimensions and facts (for example, ROW_WIDs of Dimensions and Facts, INTEGRATION_ID and DATASOURCE_NUM_ID and flags). Carefully consider which columns or combination of columns filter conditions should exist, and define indices to improve query performance. Inspect the preconfigured objects for guidance. Name all the newly created tables as WC_. This helps visually isolate the new tables from the preconfigured tables. Keep good documentation of the customizations done; this helps when upgrading your data warehouse. Once the indices are decided upon, they should be registered in the ODI Model (for more information, see Section 1.5.2, "How to add an index to an existing fact or dimension table").
Custom tables should follow the WC_ naming convention to help distinguish from preconfigured tables. Follow this procedure to add a new fact table to the Oracle Business Analytics Warehouse.
To add a new fact table:
Create the custom fact datastores and tasks. Create a WC_<fact name>_F datastore under the 'Oracle BI Applications – Fact' model. Create a WC_<fact name>_FS datastore under the 'Oracle BI Applications – Fact Stage' model. Use the WC_SAMPLE_FS and WC_SAMPLE_F datastores as templates. These datastores include all required system columns.
Note that the specific submodel that a table belongs to drives the table maintenance behavior. For example, tables in the 'Fact Stage' submodel will always be truncated during each ETL run while tables in the 'Fact' submodel are only truncated during a Full ETL run.
A fact can be defined in ODI either manually, by generating the DDL to create the table in the database or by defining the table in the database and importing the definition into ODI using the BI Apps RKM. If using the RKM, the imported table will automatically be placed in the 'Other' submodel and will need to be moved into the 'Fact Staging' and 'Fact' submodels as appropriate. The OLAP type also needs to be set for the fact table to 'Fact Table'.
To manually create a Fact Table:
In Designer, navigate to Models > Oracle BI Applications (Folder) > Oracle BI Applications (Model) > Fact Stage (Submodel), right-click the WC_SAMPLE_FS datastore and select Duplicate Selection.
Double-click the new datastore and rename it. Name and Resource Name should match the actual table name. Alias can be the same or a more user friendly value.
In the Columns subtab, add all columns.
Repeat the same steps to create the Fact Table by copying the WC_SAMPLE_F datastore under the 'Facts' submodel.
For the fact table, set the OLAP type to 'Fact Table'
Generate the DDL to create the table in the database.
To import Fact Tables into ODI:
In Designer, navigate to Models > Oracle BI Applications (Folder) and double-click the Oracle BI Applications model.
In the Reverse Engineer subtab, indicate the tables to be imported under the 'LIST_OF_TABLES' option. To import multiple tables, provide a comma separated list.
Click Reverse Engineer. A session is started that imports the table or tables into ODI.
The Reverse Engineer process places all tables in the 'Other' submodel. Drag and drop W_%_FS tables into the Fact Stage submodel and the W_%_F table into the Fact submodel.
Double-click the new fact datastore and set the OLAP type to 'Fact Table'.
Generate the DDL to create the table in the database.
Add a foreign key constraint to all dimension tables associated with this fact. The foreign key constraint ensures the Dimension SIL task is included in the generated load plan. The Dimension SDE task will be included in the generated load plan because it populates the staging table that is used as a source for the Dimension SIL task.
Drill into the Fact datastore.
Right-click the 'Constraints' subfolder below the Fact datastore and select New Reference. The naming convention is FK_<Fact Table>_<Dimension Table>. If there are multiple WID columns that need to reference the same dimension table, enumerate each with a numeric suffix. For example, FK_WC_CUSTOM_F_WC_CUSTOM_D1.
Set the Type to 'User Reference', select the dimension from the Table drop-down list and, in the Columns subtab, click the green '+' button on the top right to add a new column.
For the Foreign Table column, select the custom WID column in the fact table. For the Primary Table column, select the ROW_WID column in the dimension table.
Create an SDE and SIL task in the Custom SDE and SIL adaptor folders. Use the SDE_<Product Line Code>_SampleFact and SIL_SampleFact tasks as a template. These tasks include the logic required to populate the system columns.
Add Load Plan step to the '3 SDE Facts X_CUSTOM_FG <Product Line Version Code>' Load Plan Component.
In Designer, navigate to Load Plans and Scenarios > BIAPPS Load Plan > Load Plan Dev Components.
Navigate to SDE - <Product Line Version Code> and double-click the '3 SDE Facts X_CUSTOM_FG <Product Line Version Code>' Load Plan Component.
Select the 'X_CUSTOM_FG' step.
Click the green '+' symbol near the top right and select the 'Run Scenario Step' option.
Provide the Scenario Name, Version should be -1, Step Name should match the Task name. Set the Restart Type to 'Restart from failed step.'
Add a Load Plan step to '3 SIL Facts X_CUSTOM_FG' Load Plan Component.
In Designer, navigate to Load Plans and Scenarios > BIAPPS Load Plan > Load Plan Dev Components.
Navigate to SIL and double-click the '3 SIL Facts X_CUSTOM_FG' Load Plan Component.
Select the 'X_CUSTOM_FG' step.
Click the green '+' symbol near the top right and select the 'Run Scenario Step' option.
Provide the Scenario Name, Version should be -1, Step Name should match the Task name. Set the Restart Type to 'Restart from failed step.'
Category 3 customizations use the Universal adapter to load data from sources that do not have pre-packaged adapters.
This section contains the following topics:
Follow this procedure to add new data as a whole row into a standard dimension table in the Oracle Business Analytics Warehouse.
To add new data as a whole row into the standard dimension table:
Identify and understand the existing structure of staging tables. Refer to Oracle Business Analytics Warehouse Data Model Reference for the table structures. Non-system columns can include the null value.
Create a custom SDE interface to load the data into the staging table in the custom folder for this purpose. The staging table needs to be populated with incremental data (rows that have been added or changed since the last Refresh ETL process), for performance reasons.
Populate the INTEGRATION_ID column with the unique identifier for the record.
The combination of INTEGRATION_ID and DATASOURCE_NUM_ID is unique. Populate the INTEGRATION_ID column with the unique identifier for the record. The combination of INTEGRATION_ID and DATASOURCE_NUM_ID is unique.
After the data is populated in the staging table, use the standard SIL interfaces to populate the dimension target tables.
Each application has prepackaged logic to extract particular data from a particular source. This section discusses how to capture all data relevant to your reports and ad hoc queries by addressing what type of records you want and do not want to load into the Oracle Business Analytics Warehouse, and contains the following topics:
You can configure extract mappings and Interfaces in the Oracle Business Analytics Warehouse to accommodate additional source data. For example, if your business divides customer information into separate tables based on region, then you would have to set up the extract interface to include data from these tables.
Extract interfaces generally consist of source tables, expressions used in the target columns, and a staging table. If you want to extract new data using the existing interface, you have to modify the extract interface to include the new data by performing the following tasks:
To modify an existing interface to include new data:
Modify the existing interface to extract information from the source, and add it to an appropriate extension column.
Modify the Expressions in the target table to perform any necessary transformations.
Save the changes.
Regenerate the scenario.
You have to determine which type of extension column to map the data to in the staging table. After you modified the extract interface, you would also have to modify the corresponding load interfaces (SDE and SIL) to make sure that the extension columns that you added are connected all the way from the staging table to the target data warehouse table.
Extract interfaces (which have the SQ_* naming convention) reside in source-specific folders within the repository. Extract interfaces are used to extract data from the source system. You can configure these extract interfaces to perform the following:
Extract data from a new source table.
Set incremental extraction logic.
When you load data from a Comma Separated Values (CSV) formatted source file, if the data contains a comma character (,), you must enclose the source data with a suitable enclosing character known as a delimiter that does not exist in the source data.
Note: Alternatively, you could configure your data extraction program to enclose the data with a suitable enclosing character automatically.
For example, you might have a CSV source data file with the following data:
Months, Status January, February, March, Active April, May, June, Active
If you loaded this data without modification, ODI would load 'January' as the Months value, and 'February' as the Status value. The remaining data for the first record (that is, March, Active) would not be loaded.
To enable ODI to load this data correctly, you might enclose the data in the Months field within the double-quotation mark enclosing character (" ") as follows:
Months, Status "January, February, March", Active "April, May, June", Active
After modification, ODI would load the data correctly. In this example, for the first record ODI would load 'January, February, March' as the Months value, and 'Active' as the Status value.
To set up the delimiter for a source file:
Open the CSV file containing the source data.
Enclose the data fields with the enclosing character that you have chosen (for example, (").
You must choose an enclosing character that is not present in the source data. Common enclosing characters include single quotation marks (') and double quotation marks (").
Save and close the CSV file.
In ODI Designer, display the Models view, and expand the Oracle BI Applications folder.
Identify the data stores that are associated with the modified CSV files. The CSV file that you modified might be associated with one or more data stores.
In ODI Designer, change the properties for each of these data stores to use the enclosing character, as follows:
Double-click the data source, to display the DataStore: <Name> dialog.
Display the Files tab.
Use the Text Delimiter field to specify the enclosing character that you used in step 2 to enclose the data.
Click OK to save the changes.
You can now load data from the modified CSV file.
This section explains how to customize the way that Oracle Business Intelligence Applications loads data into the Oracle Business Analytics Warehouse.
Before you decide to enable primary extract and delete sessions, it is important to understand their function within the Oracle Business Analytics Warehouse. Primary extract and delete mappings allow your analytics system to determine which records are removed from the source system by comparing primary extract staging tables with the most current Oracle Business Analytics Warehouse table.
The primary extract mappings perform a full extract of the primary keys from the source system. Although many rows are generated from this extract, the data only extracts the Key ID and Source ID information from the source table. The primary extract mappings load these two columns into staging tables that are marked with a *_PE suffix.
The figure below provides an example of the beginning of the extract process. It shows the sequence of events over a two day period during which the information in the source table has changed. On day one, the data is extracted from a source table and loaded into the Oracle Business Analytics Warehouse table. On day two, Sales Order number three is deleted and a new sales order is received, creating a disparity between the Sales Order information in the two tables.
Figure 1-4 shows the primary extract and delete process that occurs when day two's information is extracted and loaded into the Oracle Business Analytics Warehouse from the source. The initial extract brings record four into the Oracle Business Analytics Warehouse. Then, using a primary extract mapping, the system extracts the Key IDs and the Source IDs from the source table and loads them into a primary extract staging table.
The extract mapping compares the keys in the primary extract staging table with the keys in the most current the Oracle Business Analytics Warehouse table. It looks for records that exist in the Oracle Business Analytics Warehouse but do not exist in the staging table (in the preceding example, record three), and sets the delete flag to Y in the Source Adapter, causing the corresponding record to be marked as deleted.
The extract mapping also looks for any new records that have been added to the source, and which do not already exist in the Oracle Business Analytics Warehouse; in this case, record four. Based on the information in the staging table, Sales Order number three is physically deleted from Oracle Business Analytics Warehouse, as shown in Figure 1-4. When the extract and load mappings run, the new sales order is added to the warehouse.
The primary extract (*_Primary) and delete mappings (*_IdentifyDelete and *_Softdelete) serve a critical role in identifying which records have been physically deleted from the source system. However, there are some instances when you can disable or remove the primary extract and delete mappings, such as when you want to retain records in the Oracle Business Analytics Warehouse that were removed from the source systems' database and archived in a separate database.
Because delete mappings use Source IDs and Key IDs to identify purged data, if you are using multiple source systems, you must modify the SQL Query statement to verify that the proper Source ID is used in the delete mapping. In addition to the primary extract and delete mappings, the configuration of the delete flag in the load mapping also determines how record deletion is handled.
You can manage the extraction and deletion of data in the following ways:
Deleting the configuration for source-archived records
Deleting records from a particular source
Enabling delete and primary-extract sessions
Configuring the Record Deletion flag
Configuring the Record Reject flag
Some sources archive records in separate databases and retain only the current information in the main database. If you have enabled the delete mappings, you must reconfigure the delete mappings in the Oracle Business Analytics Warehouse to retain the archived data.
To retain source-archived records in the Oracle Business Analytics Warehouse, make sure the LAST_ARCHIVE_DATE parameter value is set properly to reflect your archive date. The delete mappings will not mark the archived records as 'deleted'. For more information about extract and delete mappings, see Section 1.4.3.2, "About Working with Primary Extract and Delete Mappings".
This section contains miscellaneous information that applies to all three categories of customization in Oracle Business Intelligence Applications, and contains the following topics:
This section explains codes lookup and dimension keys.
By default, dimension key resolution is performed by the Oracle Business Analytics Warehouse in the load mapping. The load interface uses prepackaged, reusable lookup transformations to provide pre-packaged dimension key resolution. This section describes how dimension keys are looked up and resolved.
There are two commonly used methods for resolving dimension keys. The first method, which is the primary method used, is to perform a lookup for the dimension key. The second method is to supply the dimension key directly into the fact load mapping.
If the dimension key is not provided to the Load Interface through database joins, the load mapping performs the lookup in the dimension table. The load mapping does this using prepackaged Lookup Interfaces. To look up a dimension key, the Load Interface uses the INTEGRATION_ID, the DATASOURCE_NUM_ID, and the Lookup date, which are described in the table below.
Table 1-1 Columns Used in the load mapping Dimension Key Lookup
| Port | Description |
|---|---|
|
INTEGRATION ID |
Uniquely identifies the dimension entity within its source system. Formed from the transaction in the Source Adapter of the fact table. |
|
DATASOURCE_NUM_ID |
Unique identifier of the source system instance. |
|
Lookup Date |
The primary date of the transaction; for example, receipt date, sales date, and so on. |
If Type II slowly changing dimensions are enabled, the load mapping uses the unique effective dates for each update of the dimension records. When a dimension key is looked up, it uses the fact's primary or 'canonical' date to resolve the appropriate dimension key. The effective date range gives the effective period for the dimension record. The same entity can have multiple records in the dimension table with different effective periods due to Type II slowly changing dimensions. This effective date range is used to exactly identify a record in its dimension, representing the information in a historically accurate manner.
There are four columns needed for the load interface lookup: INTEGRATION ID, DATASOURCE_NUM_ID, and Lookup Date (EFFECTIVE_FROM_DT and EFFECTIVE_TO_DATE). The lookup outputs the ROW_WID (the dimension's primary key) to the corresponding fact table's WID column (the fact tables foreign key).
Dimension and Fact Tables in the Oracle Business Analytics Warehouse use the following two types of index:
ETL Index
ETL Indexes are used for Unique/Binary Tree index.
Query Index
Query Indexes are used for Non-Unique/Bit Map Index.
To add an index to an existing fact or dimension table:
In ODI Designer, display the Models view, and expand the 'Oracle BI Applications' folder.
Expand the Fact or Dimension node as appropriate.
Expand the Table in which you want to create the index.
Right-click on the Constraints node, and select Insert Key to display the Key: New dialog.
Display the Description tab.
Select the Alternate Key radio button, and update the name of the Index in the Name field.
Display the Column tab.
Select the column on which you want to create the index.
Display the FlexFields tab.
Use the settings to specify the index type, as follows:
For 'Query' type indexes (the default), define the index as an 'Alternate Key' for unique indexes and as 'Not Unique Index' for non-unique indexes.
For 'ETL' type indexes, clear the check box for the INDEX_TYPE parameter and set the value to 'ETL'. In addition, set the value of the IS_BITMAP parameter to 'N' and define the index as an 'Alternate Key' for unique indexes and as 'Not Unique Index' for non unique indexes.
Save the changes.
|
 Copyright © 2013, Oracle and/or its affiliates. All rights reserved. Legal Notices |
|
