Performance for Import + live dataset model

[matteoumbri]

I am building a composite dataset as follows:

- some Dimension tables on import mode

- a HistoryFact table which is a connection to a 1-table Power BI dataset, published on the service with incremental refresh (that's over 80 million rows)

Question is about query performances, especially on report pages with many resource-intensive visuals: 

will a report built upon that composite model perform the same, when interacted by an end-user on the service, compared to an identical report, this time built upon a full import-mode dataset? 

I am asking this because on Power BI desktop, while writing measures and testing visuals on pages, I assume it is normal that when you have all the tables loaded on cache locally, visuals load faster than when one of your main tables needs to query the data to the service over the network. 

But when a composite dataset like the one described above is published on the service, well, the whole data is already there on some Azure storage, no queries are sent to the original data sources while the end-user comsumes the report, so I would assume that the experience for the user should be as an import-mode. 

Is that correct? 

Thanks

[lbendlin · Accepted solution]

"will a report built upon that composite model perform the same, when interacted by an end-user on the service, compared to an identical report, this time built upon a full import-mode dataset? "

Yes. At that point both options are represented as Power BI Service datasets  (as opposed to pass-through Direct Query back to the data source)  so the performance is expected to be identical.