Understanding Power BI Regions, Licensing, and their limitations on Large Semantic Models
Introduction
Power BI is used by people all over the world and is thus a global product that Microsoft hosts on multiple servers around the world. This is needed for multiple reasons, one of them being the distance to the end-users/customers. When a user logs in and the Power BI server is close by, the loading times will be quicker due to a common networking factor called latency.
The different regions Microsoft uses for Power BI are:
License region
The region where your Power BI items are stored is based on your license. For unreserved capacities (Pro, PPU), the region of the Azure Tenant is chosen. You cannot change this afterward.
For reserved licenses (such as embedded), you can choose the region at the moment of buying or reserving your capacity. This cannot be changed afterward.
Closest region finder
When needing to choose a region, such as when buying a premium capacity, this tool can help you find the region closest to you:
Changing licenses / regions
One important thing to note when talking about licenses and regions is the changing of regions. When you have a dataset that is published in a certain region and want to change the workspace license to one from another region, you will be able to do so without any errors in the workspace settings. However, after changing the license for a workspace containing large model format semantic models, you'll receive this error when trying to connect to it using Power BI Desktop:
And this error in the Power BI Service:
Something went wrong: This semantic model is too big to be converted to small semantic model storage format. Make it smaller or continue using large semantic model storage format.
What happened? Well, Power BI changed the region of the workspace (e.g., from West Europe to North Europe), except for the large semantic model. These models can't be changed so easily. There are three workarounds to still change these regions:
- Republish the model
This is an easy option but not the most optimal one. Deleting and republishing a model will break all the existing connections. Depending on the number of reports using this model, you should try to prevent this in most cases.
- Turning the large semantic model format on and an off
You can turn the large semantic model format off, change the workspace license back and forth (to the desired license), and then turn the setting back on. This is an easy fix, except when the model is larger than 10GB. In that case, the option can't be turned off anymore. You can work around this by deprocessing or clearing the cube in SSMS. This will remove all the data and only keep the model information. This dropped my model's size from 13GB to 20MB. Turn off the setting, change the license, and turn it back on again. Then, reprocess the model to get all the data back in. This will cause some downtime for end-users. The model will be available but without any data during this process.
- Letting Microsoft move the model
You can request Microsoft to move the model for you. Changing this region later is possible, but has to be done by the Microsoft team and can take up to six months!
Final check
After choosing your method of choice, you can check if all workspaces use your desired workspace using the Power BI admin portal.