When you are developing a calculation view, the calculation view editor is able to highlight settings that might lead to poor performance.
To enable this, you use the Performance Analysis mode. This tool is launched from inside the calculation view editor in the toolbar (the icon looks like a speedometer). You do not need to activate or build a calculation view in order to use this tool. The basic idea is that you are able to react at design time to warnings and other information relating to your choices, which might lead to poor performance so that you can consider alternative approaches.
Performance Analysis – Detailed Information
You switch your calculation view to Performance Analysis mode by pressing the button that resembles a speedometer in the toolbar. When you do this, the Performance Analysis tab appears for all nodes except the semantics node. Choose this tab to view detailed information about your calculation view, in particular the settings and values that can have a big impact on performance.
Examples of the information presented on the Performance Analysis mode tab include the following:
The type of tables used (row or column)
Join types used (inner, outer, and so on)
Join cardinality of data sources selected (n:m and so on)
Whether tables are partitioned, and also how the partitions are defined, and how many records each partition contains
When you leave Performance Analysis mode, the Performance Analysis tab disappears from all nodes.
Performance Analysis Validation Warnings
As well as information about the calculation view, Performance Analysis mode also provides you with warnings such as the following:
The size of tables with clear warnings highlighting the very large tables
Missing cardinalities or cardinalities that do not align with the current data set
Joins based on calculated columns
Restricted columns based on calculated columns
This information enables you to make good decisions that supports high performance. For example, if you observe that a table you are consuming is extremely large, you might want to think about adding some partitions and then apply filters so that you process only the partitions that contain data you need.