Embedded Capacity Usage Consistently Runs High

[openarchdata]

We are using PowerBI embedded and do nightly refreshes of dataflows and datasets. For some reason following these refreshes we are seeing high usage (often between 40% and 60% of our capacity at A3) that remains consistent through the day. When we pause the capacity and then restart after a few minutes it often seems to correct the problem and we see our usage drop to below 10%. We've looked at Microsoft's Premium Capacity Utilization and Metrics dashboard and noticed that there appears to be a constant background process running throughout the day. However, we are unable to identify what this background process might be. Note that we have query caching turned off on our embedded capacity so it's not that.

Any ideas what could be causing this mysterious elevated capacity usage that requires a capacity pause?

[Anonymous]

Hi @openarchdata ,

Please try using the Gen2 metrics app to monitor specific usage of your capacity.

Please refer to the following document for more information.

Monitor Power BI Premium Gen2 capacities with the Gen2 metrics app. - Power BI | Microsoft Learn

Power BI embedded analytics capacity planning - Power BI | Microsoft Learn

Best Regards,

Neeko Tang

If this post  helps, then please consider Accept it as the solution  to help the other members find it more quickly. 

[jennratten]

Hello!  Background processes refer to refresh operations.  In the Premium Capacity Usage and Resource Utilization report, on the Overview page, there is a tile that shows Capacity CPU % with a breakdown by interactive and background processes.  It also includes a detail of the corresponding activities.  What do you see on this page?

[openarchdata]

Thank you for your reply. I have attached a screenshot of that tile. As you can see the background process is running high constantly. We are not constantly running refreshes though. I can't seem to determine what activity is causing that constant background CPU.

[jennratten · Accepted solution]

This visual is showing that dataflow/dataset refreshes are consuming 25-30%ish of your capacity from around 1:00am - 6:00pm.  If you hover your mouse over one of the blue bars in the chart you will see a point in time drill-through.  Click that and it will take you to a page that shows the breakdown of the operations and their consumption at that point in time.  

[openarchdata]

This is very helpful. It looks like all of our nightly refreshes are consuming capacity throughout the day. My research into this indicates that this is expected behavior because refresh CPU's are spread out over the course of 24 hours instead of all at once during the refresh process.