Uploading a 30gb dataset and sharing reports

[daveedd]

Hi all,

I work for an organisation, as a part of our O365 tenant we have:

100 Pro licences

100 Premium licenses

(both per user)

I'm looking to move more dashboarding away from other tooling and into Power Bi.  The data currently in the other toolset is 17gb, but I expect that to increase.  Total storage isn't going to be an issue with that many licenses, but our workspaces have a 10gb limit.

How can I make a model with that much data?
I had pictured creating a workspace where the model sits, which only data analysts can access, then creating other workspaces which hold the reports.

Thanks in advance

Dave

[ibarrau · Accepted solution]

Hi. I would say start data modeling. It's key to to downsize. Make sure you read some downsize or great practices for models like, only import necessary columns, make sure they have a correct data type, data model, etc.

If the data model at the end of that has a size bigger than 1gb. Then you can create a workspace and asign PPU Capacity. Then you can only share with Premium Per User users.

An alternative would be having a warehouse/lakehouse that could allow you to build PowerBi Aggregations. That composite hibrid model can let you import dimensions, aggregated facts and direct query full facts. That could work if you are kind of close to 1gb, if the model size with only the dimensions is bigger than 1gb, then don't bother.

I hope that helps,

[lbendlin · Accepted solution]

You need a Premium SKU that can hold twice the size of your semantic model. Although with Fabric that is likely to change a bit.  

10 GB is a hard limit per partition, if your data is bigger then you need to use incremental refresh and bootstrapping.

powerbi-docs/powerbi-docs/connect-data/incremental-refresh-xmla.md at main · MicrosoftDocs/powerbi-d...

[daveedd]

Thanks.  This is all so confusing and unclear.  If the partition limit is 10gb why do they advertise the premium per user model size limit as 100gb?  (bottom right)

[lbendlin]

That's a marketing/sales ploy to get you to move from PPU to Premium capacity/pro license.

Technically the model size can be bigger than the Capacity SKU memory size. But only through partitions.  I guess in the new Delta Lake world this won't matter much any more as these are automatically partitioned.

[ibarrau · Accepted solution]

Hi. I would say start data modeling. It's key to to downsize. Make sure you read some downsize or great practices for models like, only import necessary columns, make sure they have a correct data type, data model, etc.

If the data model at the end of that has a size bigger than 1gb. Then you can create a workspace and asign PPU Capacity. Then you can only share with Premium Per User users.

An alternative would be having a warehouse/lakehouse that could allow you to build PowerBi Aggregations. That composite hibrid model can let you import dimensions, aggregated facts and direct query full facts. That could work if you are kind of close to 1gb, if the model size with only the dimensions is bigger than 1gb, then don't bother.

I hope that helps,