Forum Discussion

AlexanderBP's avatar
AlexanderBP
Frequent Visitor
2 years ago

A-SKU vs AAS

Hi

 

I'm unfortunately experiencing frequent overload on my Power BI embedded service due to an increase in customer activity.


The setup is fairly simple with an Azure database, from where I have created a Power BI data model which I have embedded in our service. The data model size is around 1.5GB but it's quite complex with a few star schemas connected to each other. 

But it's still surprising to me that 5-6 people interacting with the reports simultaneously will crash an A3 SKU.

But in order to overcome this problem my thoughts are to move the data model to Azure Analysis Service (AAS) but I'm not sure if this will solve my problem. 
The cost of memory on AAS is a fraction of the price compared to an A SKU, which would make it ideal for handling reports built on my data model. My question is, if I use AAS will I then move all the querying requests from the SKU to AAS? And would that mean that I could scale my SKU down and invest in AAS instead? 
So, going from left to right in the example below:


Thanks in advance. 

Tips and Tricks

1 Reply

  • lbendlin's avatar
    lbendlin
    Super User
    2 years ago

    That should work, but it also means you are shifting all the data model development into the SSAS Tabular instance  (don't even think about implementing it as OLAP). Not sure if that is ok in your scenario.

     

     a few star schemas connected to each other

    That is a big red flag right there.  Did you know that queries across data models (star schemas in your case) can generate ginormous query text if your linking columns have high cardinality?  I have seen cases where the text of the query (just the text of the query definition) was 6 GB.  Every. Single. Query.  That can bring down your SKU in a heartbeat.