warehouse views performance in semantic model

[orlandoramirez]

Good day,

I want to know the impact of creating views in a fabric data warehouse when use them in a semantic model. We have the following scenario: We need to make a relationship from a column of a dim table to the fact table in two or more columns, for example a date column in a date-time dimension to invoice date and ship date in the fact table. Fabric only allows a relationship between two tables through one column. To solve this, we are creating views of all the dimension tables that we need to make relationships with two columns. When I add a view to the semantic model, a warning about performance issues appears in the view. We need to create several views from dim tables, and making copy of the data in other tables is not an option for us taking into account storage usage.

Since the views are mostly a select of all columns and maybe in some cases a where with a specific condition, what is the impact in performance of those views in the semantic model?

Thanks in advance.

[GilbertQ]

Yes that is correct.

[GilbertQ · Accepted solution]

Hi @orlandoramirez 

I would suggest creating a composite model of the two columns in both your fact table and your dimension table here. I will then bring this data into your semantic model and create the join on the composite model in the semantic model. This should allow a great performance within the semantic model and not having to worry about the views in the Warehouse

[orlandoramirez]

Thanks for your response. I realized I've made a mistake in my question, so I've corrected it. What I need really is to make a relationship between a column in the dim to two or more columns in the fact. Thanks for your help.