Direct Lake vs Import vs Direct Lake+Import | Fabric semantic models (May 2025) скачать в хорошем качестве

Direct Lake vs Import vs Direct Lake+Import | Fabric semantic models (May 2025)

Direct Lake + Import composite mode explained—architecture, performance tests, and where it outperforms pure Import or Direct Lake. Re‑recorded after my Fabric Conf 2025 session (the original wasn’t recorded), this walkthrough offers a critical take without the usual marketing pitch, so you can decide which storage mode truly fits your model. Download slides and read the content recap: https://sql.bi/874543?aff=yt How to learn DAX: https://www.sqlbi.com/guides/dax/?aff=yt The definitive guide to DAX: https://www.sqlbi.com/books/the-defin... 00:00:00 Intro 00:02:43 ** Import mode ** 00:09:48 Key points for Import mode 00:12:20 Demo Import mode 00:25:12 ** Direct Lake ** 00:29:29 Import vs Direct Lake 00:30:22 Key points for Direct Lake 00:33:44 Demo Direct Lake mode 00:44:27 Flavors of Direct Lake (OneLake vs SQL) 00:50:26 ** Design considerations ** 00:51:04 Semantic model design 00:51:51 Organizational issue 00:54:00 Views on Lakehouse 00:57:00 ** Composite models ** 00:57:27 Two separate models (DQ over AS) 00:59:20 Demo composite model (DQ over AS) 01:04:45 Key points for DQ over AS 01:05:19 ** Direct Lake + Import ** 01:08:07 Demo Direct Lake + Import 01:29:18 Key points for Direct Lake + Import 01:31:05 Conclusions Additional information, corrections, and notes. 1. At 13:38, the option "Connect to SQL endpoint - DirectQuery" is just a regular DirectQuery connection, not DirectLake on SQL, as mentioned in the video. Currently, Microsoft does not plan to support Direct Lake on SQL in Power BI Desktop. 2. At 36:28, data previews don't show for Direct Lake when creating relationships, no matter how long you wait. The logic in Power BI Desktop tries to guess the direction of a relationship based on row counts. It will pick many sides based on higher row counts and one side based on lower row counts between the two tables. 3. At 38:35, you don't have to go to the cloud to create a report. You can do it in a new instance of Power BI Desktop. You can also go to the name of the file in the header to quickly navigate to the published semantic model in the service. 4. At 1:01:19, the relationship can also be created by right-clicking the "Relationship" node in the model explorer, which can help bypass the slow dialog bug. 5. At 01:18:29, the refresh will fail the first time and then clear the cache. You can immediately do it a second time, and it will work. There is no need to just "wait" as suggested in the video.