Using Direct Lake for data sources
In Chapter 5, Direct Lake semantic models were introduced along with performance tuning in Import Mode and DirectQuery. It is now the third option for a semantic model. The performance of these different types of semantic models is created for a specific scenario. In the case of Direct Lake, the option is for large data volumes. For best results over Import Mode, as of the printing of this book, there is a need to have 100s of millions or billions of rows and 100+ GBs of data for Direct Lake to exceed the performance of Import Mode. We expect Microsoft to improve on this performance since Fabric was made GA in November of 2023.
The other requirement for using Direct Lake instead of Import Mode would be near real-time data in the model. Direct Lake is the replacement for DirectQuery on big data sources. Import Mode requires a data refresh when new data is available. Direct Lake does not need a data refresh because the connection is to the Delta...
