Large models in Power BI Premium (preview)

Power BI datasets can store data in a highly compressed, in-memory cache for optimized query performance to enable fast user interactivity over large datasets. The large models feature allows datasets in Power BI Premium to grow beyond 10 GB in size. The size of the dataset is instead limited by the Power BI Premium capacity size, which is similar to how Azure Analysis Services works in terms of model size limitations. For more information on capacity sizes in Power BI Premium, see Capacity nodes. You can set up large models for all Premium P SKUs and Embedded A SKUs; but they work only with the new workspaces.

Large models do not affect the PBIX upload size, which is still limited to 10 GB. Instead, datasets grow beyond 10 GB in the service on refresh. You can use incremental refresh to configure a dataset to grow beyond 10 GB.

Enable large models

To create a dataset that grows beyond 10 GB, follow these steps:

  1. Create a dataset in Power BI Desktop and configure an incremental refresh.

  2. Publish the dataset to the Power BI Premium service.

  3. Enable the dataset for large models by running the PowerShell cmdlets below. These cmdlets cause Power BI to store the dataset on Azure Premium Files and not to enforce the 10-GB limit.

  4. Invoke a refresh to load historical data based on the incremental refresh policy. The first refresh could take a while to load the history. Subsequent refreshes should be faster because they are incremental.

PowerShell cmdlets

In the current version of large models, enable the dataset for Premium Files storage using PowerShell cmdlets. You must have capacity admin and workspace admin privileges to run the PowerShell cmdlets.

  1. Find the dataset ID (GUID). On the Datasets tab for the workspace, under the dataset settings, you can see the ID in the URL.

    Dataset GUID

  2. From a PowerShell admin prompt, install the MicrosoftPowerBIMgmt module.

    Install-Module -Name MicrosoftPowerBIMgmt
  3. Run the following cmdlets to sign in and check the dataset storage mode.

    (Get-PowerBIDataset -Scope Organization -Id <Dataset ID> -Include actualStorage).ActualStorage

    The response should be the following. The storage mode is ABF (Analysis Services backup file), which is the default.

    Id                   StorageMode
    --                   -----------
    <Dataset ID>         Abf
  4. Run the following cmdlets to set the storage mode to Premium Files and check it. It can take a few seconds to convert to Premium Files.

    Set-PowerBIDataset -Id <Dataset ID> -TargetStorageMode PremiumFiles
    (Get-PowerBIDataset -Scope Organization -Id <Dataset ID> -Include actualStorage).ActualStorage

    The response should be the following. The storage mode is now set to Premium Files.

    Id                   StorageMode
    --                   -----------
    <Dataset ID>         PremiumFiles

You can check the status of dataset conversions to and from Premium Files by using the Get-PowerBIWorkspaceMigrationStatus cmdlet.

Dataset eviction

Power BI uses dynamic memory management to evict inactive datasets from memory. Power BI evicts datasets so it can load other datasets to address user queries. Dynamic memory management allows the sum of dataset sizes to be significantly greater than the memory available on the capacity, but a single dataset must fit into memory. For more info on dynamic memory management, see How capacities function.

You should consider the impact of eviction on large models. Despite relatively fast dataset load times, there could still be a noticeable delay for users if they have to wait for large evicted datasets to be reloaded. For this reason, in its current form, the large models feature is recommended primarily for capacities dedicated to enterprise BI requirements rather than capacities mixed with self-service BI requirements. Capacities dedicated to enterprise BI requirements are less likely to frequently trigger eviction and need to reload datasets. Capacities for self-service BI on the other hand can have many small datasets that are more frequently loaded in and out of memory.

Checking dataset size

After loading historical data, you can use SSMS through the XMLA endpoint to check the estimated dataset size in the model properties window.

Estimated dataset size

You can also check the dataset size by running the following DMV queries from SSMS. Sum the DICTIONARY_SIZE and USED_SIZE columns from the output to see the dataset size in bytes.

 [DATABASE_NAME] = '<Dataset Name>') //Sum DICTIONARY_SIZE (bytes)

 [DATABASE_NAME] = '<Dataset Name>') //Sum USED_SIZE (bytes)

Limitations and considerations

Keep in mind the following restrictions when using large models:

  • Bring your own key BYOK encryption: Datasets enabled for Premium Files are not encrypted by BYOK.

  • Multi-geo support: Datasets enabled for Premium Files will fail on capacities where multi-geo is also enabled.

  • Download to Power BI Desktop: If a dataset is stored on Premium Files, downloading as a .pbix file will fail.

  • Supported regions: Large models are supported in all Azure regions that support Premium Files Storage. To learn more, see Products available by region, and consult the table in the following section.

Availability in regions

Large models in Power BI are only available in certain Azure regions that support Azure Premium Files Storage.

The following list provides regions where large models in Power BI are available. Regions not in the following list are not supported for large models:

Azure region Azure region abbreviation
Australia East australiaeast
Australia Southeast australiasoutheast
Central US centralus
East Asia eastasia
East US eastus
East US 2 eastus2
Japan East japaneast
Japan West japanwest
Korea Central koreacentral
Korea South koreasouth
North Central US northcentralus
North Europe northeurope
South Central US southcentralus
Southeast Asia southeastasia
UK South uksouth
UK West ukwest
West Europe westeurope
West US westus
West US 2 westus2

Next steps

The following links provide information that can be useful for working with large models: