Dataflows, multiple workspaces, and Power BI Embedded - where to assign capacity?

[bheepatel]

Hi everyone

I have a scenario where I have the following three Workspaces:

• Workspace A: I have multiple Dataflows in this workspace reading data from SQL Server DB that refresh daily.
• Workspace B: I have a single report and a single dataset in this workspace. The dataset reads data from the Dataflows in Workspace A and this dataset also refreshes daily.
• Workspace C: I have multiple reports here which makes use of the dataset from Workspace B (using the Shared Dataset concept).

I now want to embed the reports from Workspace C in our own portal using Power BI Embedded (embed for your customers solution).

I have two questions:

1. If I purchase the relevant capacity, which workspace do I assign it to? Do I need to assign capacity to all of them? If yes, how would I know how much to assign to each workspace?
2. Is there a better architecture than above? It is set out in the current way because we need to separate the access to the Dataflows and the dataset to different people within the organization.

Thanks in advance 🙂

[Tutu_in_YYC]

Ibendin has already address the complexity of this setup. You can still continue with this if you need to.

If you have power bi premium, you can simply assign all the workspaces with Premium capacity to make things easy. You cant really customize how much capacity you assign to a workspace, it is basically either you have it or not.

But theoretically you only need the workspaces that has the dataset and the report to be on premium capacity, i.e B and C. Because Power BI embedded analytics requires dataset id and report id, and they can be from different workspaces. And the service principal has to be in both B and C workspaces.

[lbendlin]

that is a very unfortunate setup.  The dataset chaining issues are already bad enough, but now you add dataflows and embedded into the equation too.  It's as if you are trying to find the most complex setup  imaginable.

Maybe you can revisit your requirements and combine workspaces A and B?

Think of this in terms of "Import Mode" versus "Direct Query Mode".  The dataflows in A are in Import Mode - nothing you can change there. The datasets in B are in import mode too - again nothing you can change. The datasets in C are in Direct query mode, which means they don't really consume resources in C.  So you could make C really skinny but need to make B rather beefy.  We haven't yet figured out how dataflow size actually reflects on a capacity. Documentation on that seems to be scant.

[bheepatel]

Hi @lbendlin 

Thanks for your input. We have the setup in A because other users make use of the dataflows for other reports in other workspaces.

To make it less complex, I can combine B & C so they are in a single workspace. Would I still need to assign capacity to A or can I get away with assigning it solely to the single workspace (B+C)?

[lbendlin]

The idea of a shared ("golden")  dataset is that you prepare the data once, and then use it against many different audiences.  You should combine A+B, not B+C.