PowerBI embedded for 250K users to use 30 odd reports

[Subrat]

Hello All,

I have a use case for around 250K customers outside the organization to use 30 odd reports of high data volume(GB) to use as async download mechanism. Is PowerBI embedded the right approach? if so what are the recommendations?

Thanks

[Don-Bot · Accepted solution]

Why do you need to add them through AAD?  Or do you mean you would have a service principal setup in AAD that would handle creating the embedded tokens?  

Ours is setup so the workspaces/reports created permissions are given to a service principal.  Which is what is used to generate the tokens for embedding.  RLS is also used to filter things down by user.  

The number of models doesn't matter.  You can have as many as you want.  Maintaining it might be annoying but there are ways to handle that.  

The other thing to consider is the capacity size of what you want to accomplish.  Which corresponds to the size of the models you are using.

[Anonymous · Accepted solution]

Hi,Don-Bot ,thanks for your concern about this issue.

Your answer is excellent!
And I would like to share some additional solutions below.

Hello,@Subrat . I am glad to help you.

Don-Bot has provided excellent help in laying out various aspects of the do's and don'ts of using PowerBI Embedded.
Here are some additions I made to the suggestions he made.
1. Use Azure Active Directory

It is common practice to use AAD to manage users, and using Service Bodies simplifies this process and ensures secure generation of embedded tokens. Row-level security (RLS) can also be achieved through AAD. However, there are some limitations that you need to be aware of.
URL:
Generate an embed token in Power BI embedded analytics - Power BI | Microsoft Learn
 

2. Management of multiple models/reports.
Manually managing multiple semantic models is difficult, but automation tools and scripts can simplify management, and the use of service bodies can ensure security and simplify user management.
Using the Rest API is a good option, and you can use them in combination
The XMLA EndPoint mentioned by Don-Bot is also a very powerful tool. With XMLA EndPoint, you can use a variety of client applications and tools
(e.g. Tabular Editor, SSMS, etc.) to manage and extend the Power BI semantic model.
Of course all of these functionality options need to be built into the SKUs you choose.

If you have a small number of users and not a lot of concurrent access, you can choose a lower capacity SKU (e.g. A SKU).
For large scale users and high concurrent access, it is recommended to choose a higher capacity SKU (e.g. P SKU) to ensure performance and responsiveness.

You need to choose the right solution for your team to achieve the most efficient use of resources.

And I hope the following documents are helpful.
URL:
Power BI embedded analytics documentation - Power BI | Microsoft Learn
Capacity and SKUs in Power BI embedded analytics - Power BI | Microsoft Learn

I hope my suggestions give you good ideas, if you have any more questions, please clarify in a follow-up reply.

Best Regards,

Carson Jian

[Subrat]

Thanks Don for your response.

The recommendation around whether to go for an API based approach or PowerBI embedded? 

These 30 reports will have 30 different semantic models and will have to add the customers through AAD( not sure if there's any challnege) to be able add those customers to AAD.

[Don-Bot · Accepted solution]

Why do you need to add them through AAD?  Or do you mean you would have a service principal setup in AAD that would handle creating the embedded tokens?  

Ours is setup so the workspaces/reports created permissions are given to a service principal.  Which is what is used to generate the tokens for embedding.  RLS is also used to filter things down by user.  

The number of models doesn't matter.  You can have as many as you want.  Maintaining it might be annoying but there are ways to handle that.  

The other thing to consider is the capacity size of what you want to accomplish.  Which corresponds to the size of the models you are using.

[Don-Bot]

What sort of recommendations are you looking for?  And what kind of setup are you planning?

For instance, how big will your main semantic model be?  Will the 30 reports run off of 1 model or 30 different models?

We have an apps own data embedded setup.  WE don't have 250k customers so our footprint is a lot smaller than yours but embedded works wonderfully for us.  

We have 4 main semantic models.  That are shared across 400+ clients embedded in our website.  With XMLA Endpoint there are a lot of flexibility options that helps.