Hi all,
Native HANA-setup: I have built a Calculation View with a Star Join.
Very basic: A fact table and a number of dimensions.
When dimensions are joined to the fact table I have used referential joins.
I have observed the following:
When I do a select distinct on a dimension value the query results in a plan that joins the dimension to the fact-table. As my fact-table currently holds +500 mio. rows and furthermore is created on several joins it requires a lot of resources. The result is however very nice, as it gives me dimension-values that holds data only. But it is ok to show all dimension-values even though there is no data. So I tried to change the join type in the star join for one of the dimensions to a left-outer join. That result is fine also as it does not force a join to the fact table and hence the query executes very fast and gives me a list of unique dimension values.
But it seems as when I change more joins from referential to left-outer the results does not persist. As soon as two dimensions are joined to the fact table using left-outer joins it starts joining towards the fact table again and the performance decreases. Are there any modeling issues that I have not thought of, or what is the reason behind this behaviour?
BR,
Erling