Power BI Dataset Scale-Out: Better Performance for a Price

• Power BI Dataset Scale-Out (in preview) creates read-only dataset copies (replicas) to improve performance.
• It can help heavily used datasets that need isolation for data refreshes or that supply data to thousands of users.
• Each dataset copy doubles the quantity of resources and storage used, which may impact capacity and tenant limits.

Dataset Scale-Out is a new Power BI Premium load balancing feature in preview, designed to help enterprises resolve common performance issues that happen with large-scale applications. The new feature is only available for Power BI Premium deployments, as the Power BI Pro environment is not designed for such large-scale usage. The feature requires new forms of management and can increase cost if not controlled properly.

The Problems It Solves

A Power BI Premium capacity is configured with a set number of virtual cores (vCores) that are allocated automatically by the service to support different operations, such as dataset refreshes, data queries, report rendering, and user interactive features. For example, when users open a report and the report queries a dataset, the service automatically creates a container using some of the capacity’s vCores to hold the dataset so that it can service user requests. The number of vCores and amount of memory is determined automatically by the service, and customers cannot request more resources for a specific dataset.