Reading the documentation for datafusions:
DataFusion: Advantages and Disadvantages
Because a DataFusion is a just a view of the data, the output doesn’t have to be materialized and indexed into the Domo databases, which makes it so that data can be combined, visualized, and updated very quickly in Domo.
It makes data transformations fast when you’re working with a lot of data; more specifically, any DataSets with more than 20 million rows and >25 columns of data should be combined using a DataFusion.
You can leverage creative Beast Mode calculations within Analyzer to specify which data you want to display and how.
You can control naming of columns during initial setup of the DataFusion.
You can control which columns to include or exclude at any time.
You can only do joins or appends using DataFusion.
You cannot use DataFusions as inputs for DataFlows.
You have less backend control over renaming or editing your column names after the initial creation of the DataFusion.
The first bullet point uses the word "view". To an Oracle developer it sounds as if the row count in the fusion would not actually count against the warehouse total row count (hints being "VIEW", "MATERIALIZED" and "INDEXED INTO DOMO"). When I create a 35M row fusion, my warehouse count increased by 35M. Can anyone explain?
Even though the data has not been indexed every time the fusion runs it still has to store the data somewhere or you wouldn't be able to visualize the data in cards, so the data has 35M rows cached until the next time the datafusion runs.
But being stored (indexed) does not agree with the first bullet point:
To me, this says its virtual.