Forum Discussion

tessahurr's avatar
tessahurr
Icon for Microsoft Employee rankMicrosoft Employee
5 years ago

Share your thoughts on DirectQuery for Power BI datasets and Azure Analysis Services (preview)

Hit Reply and let us know what you think of the DirectQuery for Power BI datasets and Azure Analysis Services.  To learn more about this feature, please visit this blog post or our documentation.   ...
General Comment
Mariusz's avatar
Mariusz
Icon for Community Champion rankCommunity Champion
5 years ago

Hi tessahurr 

 

First of all thanks for this amazing future, I have seen the demo ( during MBAS 2020... I think ) and was waiting for the preview with anticipation.

I do have a few questions I hope you will be able to answer: 

While connecting to two datasets Sales and Order ( both star schemas ) and both datasets/models contains the same dimension Product: 

  • How do I remove one of these dimensions?
  • Will it make a difference in performance if I build relationships between two different datasets ( Product from Sales dataset  filtering Orders fact from orders dataset )
  • Will you be able to merge dimensions if identical?
  • Will this stay as Pro or will this be a premium only future when generally available?

Many Thanks 

Mariusz

tessahurr's avatar
tessahurr
Icon for Microsoft Employee rankMicrosoft Employee
5 years ago

1. Can you give more info on the data sources you are connected to for Product and Product 2?

2. I don't believe so, but perhaps jeroenterheerdt can help answer your question about performance.

3. This is an interesting idea. We don't allowe for that right now, but can you describe a bit more about your expected behavior?

4. When this feature is GA it will be available for both shared and premium.

 

Thanks!

  • jeroenterheerdt's avatar
    jeroenterheerdt
    Icon for Microsoft Employee rankMicrosoft Employee
    5 years ago

    tessahurr I am not sure who jeroent is, but that's not me 🙂 regarding the performance question from Mariusz: if I understand it correctly you are asking if there is performance difference between filtering within one source compared to across two sources. While the answer is it depends, I would expect the performance for filtering across two sources to be less than within the source, unless one of the tables is significantly larger than the other.

  • nicg's avatar
    nicg
    Icon for Advocate II rankAdvocate II
    5 years ago

    I have the same question as Mariusz, and I think it gets at architectural issues that Power BI has with conformed dimensions. It's never been clear to me the best way to handle them as there is no concept of shared dimensions like there used to be in SSAS MD.

     

    If there's two models, both with a copy of the same Product dimension and those two models are "composited" together, you're going to have two Product dimensions. How do you get the two models to work together across this dimension? Can you remove one? Can you link them together somehow? It seems to me that each model may contain its own measures with references to its own Product dimension. How do we keep these working?

    • Anonymous's avatar
      Anonymous
      Not applicable
      5 years ago

      Well, is this realy related to "Composite"? Same question with "Import"...
      The usual way is to have just one dim with several relations and to build measures with "USERALTIONSHIP".

      • nicg's avatar
        nicg
        Icon for Advocate II rankAdvocate II
        5 years ago

        Hi Oliver,

         

        I'm not sure I follow you. I'm talking about multiple models, not multiple tables within a single model.

         

        Model A has a star schema that includes Product dimension.
        Model B has a star schema that includes Product dimension.

        Model C is a composite of A and B. C now contains Product(A) and Product(B). What's the right way of handling this situation?

    • jeroenterheerdt's avatar
      jeroenterheerdt
      Icon for Microsoft Employee rankMicrosoft Employee
      5 years ago

      conforming dimensions is something we are still working on.