Skip to main content

Announcing general availability for composite models on Power BI Datasets and Analysis Services models

Headshot of article author Jeroen ter Heerdt

In December 2020, we launched the preview of DirectQuery for Power BI Datasets and Analysis Services. Since then, we have worked with many of you to improve on this feature and get it ready for general availability. A couple of highlights of the things we did along the way:

  • Support for SQL Server Analysis Services Tabular models (version 2022 required).
  • Made sure display folders, sort by column properties and format strings are persisted.
  • Introduced sub setting, and the ability to connect to perspectives.
  • Consumers can now view a report built using this feature with Read permissions / Viewer role instead of Build permission / Contributor role. More about that later in this blog.

To simplify the name of the feature, starting today we will refer to models built using this feature as composite models. We will drop the name “DirectQuery for Power BI Datasets and Analysis Services.” That name served its purpose during the preview period to be able to clearly identify that this was a preview feature. However, you are making a composite model on top of sources, which happen to include these “special” sources, such as Power BI Datasets.

Announcing general availability

Today, we announce that the preview is officially over and that composite models based on Power BI datasets and Analysis Services Tabular models are now generally available and fully supported on Premium, PPU and new Pro workspaces.

We will announce general availability for existing Pro workspaces later as we finish the backend changes mentioned later in this blog post. Please keep in mind that as these back-end changes are happening, consumers might no longer need Build permissions to view a report built using this feature, but instead they can rely on just Read permissions. When this happens, you can safely take away the Build permissions for your consumers.

If you are currently using Pro workspaces and want to switch over to the Read permission model right away, you can. However, it requires you to create new workspaces and host the datasets and reports there. As a reminder, going forward you will need to enable the ‘Allow XMLA endpoints and Analyze in Excel with on-premises datasets’ setting enabled in your tenant to use this feature.

Required permissions to view a report

When we started the preview, any consumer of a report that leveraged a composite model based on a Power BI dataset was required to have Build permissions or the Contributor role. Since then, we have changed this for Premium and PPU workspaces: readers of reports based on datasets in Premium or PPU workspaces just require Read permissions (or the Viewer role).

However, for most Pro workspaces Build permissions are still required in the scenario above. We are making backend changes to align the Pro workspaces with the Premium and PPU workspaces so everyone consuming these reports will just require Read permission or the Viewer role, regardless of the workspace type the data is stored in.

As you can imagine, these backend changes are not easy and take a lot of time. In fact, these changes are still underway and take longer than we expected.

What this means for you is:

  1. If you only use Premium / PPU workspaces only Read permission is required to view a report.
  2. If you use a mixture of Premium / PPU and Pro workspaces or use Pro workspace exclusively, some of your reports might currently only require Read permissions while others will still require Build. The ratio of these will shift over time to the vast majority just requiring Read as we complete the backend changes.
  3. If you currently use Pro workspaces, some datasets will not benefit from the backend changes and will still require Build permissions, even after the back-end changes are completed. This includes datasets that:
    1. use one or more unsupported sources, and/or
    2. consume a large amount of memory, which is true for about 0.03% of all datasets.

A reminder about unsupported sources

We have seen quite a few people making DirectQuery connections to unsupported sources, such as usage metrics and real-time data sets. While during the preview you might not have received an error when using these sources, you will start seeing an error going forward as we tighten up all loose ends. Even if you do not see an error, you should still use this feature with supported sources. The unsupported sources are documented, so please make sure you use supported sources for your reports that rely on this feature now.

More information

Read more about composite models in our documentation. Also, please read the recommendations in our guidance information.