Power BI Performance Tables vs Flat Files

[asjones]

In general does Power BI Service (online web or Desktop) do better performance wise with a as single (large flat file) or multiple relations (star type schema etc.).

I was on a webinar and the presenter said the way Power BI was it do better performance wise with a star schema relationships as opposed to a single super large flat file.

When the answer is “it depends” what factors are there to consider (outside of measures vs calculated columns and optimizing DAX).

I ask specifically as I know people with competing products that say performance with a single large flat file is so much better than dealing with relationships.

Personally I like the relationships, as it can simplify the process. However as the volume of data increases what is beter?

Thanks

Alan

[Greg_Deckler]

This is probably one of the better articles on the subject:

https://powerpivotpro.com/2016/02/data-modeling-power-pivot-power-bi/

[asjones]

thanks... i have seen that article before.... Much of it reminds me of good database design prinsicples (ER Diagrams etec). The one area that talked about "performance" said

    * Wide Flat Tables = Poor Performance
    * Narrow Data Tables with separate Lookup Tables (Star Schema) = Improved Performance
I don't disagree, but have also learned in reality that depennding how the software is implemented.  It could be better performance wise to be flat. I have bene told Tableu does better with flat files and not relations.

My lookups in a relatin would remove some columlns, but would not reuce the number of rows my primary tables.

I guess I was curious if anyone had tested and done any simple comparisions on speed.

thanks for the response... if anyone has more detial would love to see it out of curiosity.

thanks

Alan

[Anonymous]

@asjones - did you find anything out through your investigations?
I'm querying the same question and I'm unsure of the answer too.