!--a11y-->
Aggregates and HPA Indexes 
Relational aggregates and high performance analytics (HPA) allow you to improve the performance of BI queries when reading data from an InfoCube. These are redundant data stores in a BI InfoCube. In relational aggregates, data is stored in an aggregated form; in an HPA index, data is compressed but not aggregated.
Relational aggregates are useful if you want to improve the performance of one or more specific BI queries, or specifically improve the performance of reporting on characteristic hierarchies.
HPA is particularly useful in cases where relational aggregates (or other BI-specific methods of improving performance such as database indexes) are not sufficient, are too complex, or have other disadvantages.

For example, if you have to maintain a large number of aggregates for one particular InfoCube, you can use HPA to avoid this high maintenance effort.
Integration
You can use relational aggregates and an HPA index for the same InfoCube. A BI query always tries to use performance-optimized sources and checks the sources from which it can draw the requested data in the following order:
...
1. OLAP cache
2. HPA index
3. Relational aggregates on the database
4. InfoCubes on the database
When the query is executed, it is clear to the user whether data is being read from an aggregate, an HPA index or an InfoCube.
In each maintenance transaction, you can temporarily deactivate one or more aggregates or the HPA index in order to test the aggregate or HPA index, for example, to measure performance or analyze data consistency.
You can also
execute the relevant query in the query monitor (transaction RSRT)
using a corresponding debug option: Choose
Execute + Debug. In the Debug Options dialog box, choose:
· Do Not Use Aggregatesto execute the query with an InfoCube, as long as no HPA index exists
· The Do Not Use HPA Index option to execute the query with aggregates or an InfoCube
Constraints
In contrast to relational aggregates, HPA indexes are only possible with InfoCubes that have cumulative key figures and not with InfoCubes that have non-cumulative key figures.