NOTE- Please refer dimensional modeling for the concepts
An OLAP tool should be able to allow you:
- Change in the dimension attributes
- Change in the dimension hierarchy paths
- Change in the FACTS-measures
- Removing OR adding the instances across different levels in the dimensional hierarchy
The changes can be from being:
- Superficial (typically adding, updating OR removing a record in the dimensional model), to
- Fundamental (changing a dimension structure).
If the impact is limited to the re-calculation of the cube in OLAP, without any back-ward impact on Data Warehouse and ETL, we may call it a tactical change. However, if the change requires the re-calculation of Data warehouse, and also the re-programming ETL process, it is more fundamental.
For Up-Stream Dimensional Model Change Impacts
An OLAP tools should be able to handle tactical changes with:
-
Adequate tracking: Keep an audit trail of the changes done, and also to maintain the versions track of pre and post changes.
- Adequate propagation with-in the cubes on the impacts: If you are changing the level in a hierarchy path, (Say, an additional level of sales manager between sales head and sales team-leader), the cube need to be re-calculated wherever the instances of this level are used in the multi-dimensional arrays of MOLAP.
- Management of cache: Given the change in the cubes, the cache at various layers, will need to be reset-flushed completely OR selectively.
Coming to more fundamental changes (the change impacting the Data Warehouse data and ETL), an OLAP should be able to:
-
Use common meta-data to maintain the track of changes of meta-data from the source system to staging area to Data-Warehouse.
- Propagating the impacts of the changes to Data warehouse dimensional model, ETL routines and also any changes needed to make in the source systems.
For example, addition of a sales manager level (assuming that Data-Warehouse database did not have data related to the Sales Manager level), the ETL routines will need to change to get the data from the source system (assuming the data exists in the source system).
In case the source system does not have the data. It means that you will not find a way to capture the data. Essentially an OLAP server (along with the central data management) should be able to identify all the upstream impacts.
For Down Stream Dimensional Model Impacts:
A re-calculation of cube due to the changes in dimensional and Measures-facts structure, should be handled in an ideal OLAP solution
-
By identifying the queries which are linked to this change.
- Helping the management of pre-existing reports and views: You may need to re-run the reports OR views for previous period (For example, after adding the sales manager level, you may like to re-run the reports for all the months since the beginning of the year). The tracking of the reports and other outputs using the end-user tools (like Enterprise Reporting Tools and Analytics tools).
|
