The recommended way to develop high-performance cubes is as follows:
- Define a cube schema without any calculated elements, that is without calculated measures, calculated members, or named sets.
- Publish the schema to the Teradata BI Model Repository using Teradata Schema Workbench.
- Use the Teradata OLAP Connector User Guide (B035-4105) to set up your chosen BI client application to test the specified cubes defined in the cube schema.
- Add AJIs to significantly increase query performance. Use the Teradata Database Explain feature to verify you are actually hitting the AJIs.
- Create a calculated measure for a cube and test it. If the calculated measure does not generate the correct value, check to ensure the MDX expression is logically correct. To check if the MDX expression is semantically correct, see MDX Validation Using MDX Sample.
- Where possible, use MDX functions, methods, and operators that are calculated by the underlying Teradata Database instead of using functions, methods, and operators where the Teradata OLAP Connector needs to retrieve all the data from the Teradata Database and do the calculations on the client system. The list of functions, methods, and operators which can be calculated in the Teradata Database are listed in Supported MDX Functions and Operators.
- Use your Teradata Database Explain feature to verify the calculated measure is hitting the AJIs as expected. If the calculated measure is not hitting the AJIs, recheck the MDX expression of the calculated measure and adjust the AJIs.
- Add custom named sets or calculated members. Repeat steps 5 to 7 to ensure the MDX expression is correct, and verify that the named set or calculated member is actually hitting the AJIs.