Teo Lachev digs into a preview feature:
Microsoft announced a public preview of Power BI Dataset scale-out (DSO) for Power Premium, Premium per User (PPU), and Power BI Embedded. In the comments below the announcement, the article implies that this feature is a replacement for the Azure Analysis Services scale-out. “If you have an AAS scale out and you migrate your databases (aka models aka datasets aka cubes) to Power BI Premium, you get scale out automatically and at no extra cost.” Scaling out for free? Sure, where do I sign?
But then further down the comments, we have this clarification “[Power BI DSO happens] if a dataset is on peak load and the vcores of your capacity aren’t maxed out. Keep in mind that scalability on a single instance isn’t linear. By scaling out, we can achieve a better utilization of available CPU resources for high workloads. On the other hand, if your vcores are already maxed out, then scaling out brings no further perf benefit.” Confused? So was I, and I reached for clarification to Microsoft. Below, is my best understanding of what happens behind the scenes.
Click through for the answer.