Power BI dataset connecting Dataflow - shared capacity usage and parallel refreshes

[MiraNi]

We have datasets connecting to dataflows. Some dataflows include common dimensions, which are used in almost every dataset. I want to understand, how is the connection from dataset to dataflow meaning csv file in Azure Data Lake working - is it fast? What about when more than one dataset is connecting to same dataflow at the same time, how is the capacity shared and should this be avoided if making refreshes slow? 

I tried to find documentation about this but didn't succeed. 

[itchyeyeballs · Accepted solution]

Each dataset that uses the dataflow will still need to be refreshed individually, think of it like a centrally stored csv file. It will be faster than a locally stored version.

If the quantity of data in the dataflow is large you should probably stagger the refreshes of the datasets that use it.

[ToddChitt]

My understanding of Dataflows is that Power BI stores the data in a proprietary format in some blob storage it provisions for you behind the scenes. I think there is also the option to provision your own and have Power BI use that. So when the Dataflow gets refreshed, it is pulling data from some slow source, like your on-premise CSV file, and storing it in some location and format it can access easily. 

There should be minimal considerations of refreshing multiple Power BI models that all consume the same Dataflow.

I do not know how it affects capacity metrics, but since the data is technically not stored in Power BI, it may not affect it at all. 

I have seen our clients provision one Workspace that just contains Dataflows and nothing else, and that seems to work well.

Any experts out there, please feel free to correct my assumptions.

Regards.