CPU Consumption Dataflows inconsistency

[Anonymous]

Hello,

I've remarked the following inconsistency regarding the CPU consumption of the dataflows:

In the Health metris the average consumption for the last 7 days is very low:

Meanwhile in the Capacity Metrics for the same capacity the workload is very high, clearly the average should be more than 7%:

The average duration of a dataflow is 2 minutes, on average there are 2 refreshes of a dataflow

Do you know what is the reason of this difference? What maybe causing those spikes?

[v-xicai]

Hi @Anonymous ,

CPU usage should more related to your custom merge steps, you can try to remove merge steps to confirm if this issue disappears. For reference: Creating and using dataflows in Power BI .

To improve the performance, if you have a Premium license, you could follow this blog to set dataflow container size: https://blog.crossjoin.co.uk/2019/04/21/power-bi-dataflow-container-size/ . If you have a Pro license, you could try the following workarounds:

1. You could broke up a big model into entities in 5-6 different dataflow in the same workspace. Then stagger the refresh time.
2. You could reduce the queries from that biggest table.

There are several reasons and scenarios to cause the difference, and you may refer to the articles:.

Power BI Premium Metrics app

Monitor capacities in the Admin portal

Best Regards,

Amy 

Community Support Team _ Amy

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

[Anonymous]

Hello,

Thanks for your ideas.

Please keep in mind the dataflows are executing in 2 minutes. Additionally they are very small and already well-optimized.

All of them are running on the dedicated 12v-core capacity.

The question: is  why there is a difference between health metrics and Capacity Metrics App? What causes the inconsistent and which measurer should be trusted?