Low performance on relationship between 2 fact & 2 dim tables (non star schema)

[Anonymous]

Hi all,

We have the following problem. This is a very small part of our model. Unfortunately in our real scenario we could not impelemnt the star schema in every area, due to circular dependencies between tables. Thus, we came up with the following schema which allows us to create all the necessary reports.

The case is that when I insert columns/measures from 2-3 tables on a visual in Power BI, the performance is ok (typically responds within a few minutes). When I insert columns/ measures/ filters from all the above tables the performance is about 30-40 mintues slower, which is unacceptable from our client. The profiler shows that the query creates multiple (AROUND 100) small temp tables without any obvious reason.

Does anyone knows what we may do or why this is happening?

Thanks a lot,

TP

[Anonymous]

It's true what you've mentioned. However, it is the only way to apply filtering from all the DIM tables at the same time.

So, you believe there is no other way to tackle the performance issue?

[AlexisOlson]

"it is the only way to apply filtering from all the DIM tables at the same time"

I doubt this is true. It may be the easiest way but if you're having performance issues, then you may need to re-evaluate.

[AlexisOlson]

It looks like you have bidirectional relationships between multiple fact tables. This is often very expensive computationally and should generally be avoided.

Recommended reading:

https://www.sqlbi.com/articles/different-options-to-model-many-to-many-relationships-in-power-bi-and...

https://www.sqlbi.com/articles/bidirectional-relationships-and-ambiguity-in-dax/