A dashboard connects to a SQL warehouse, not to a notebook cluster
On the canvas
Dashboard → SQL warehouse → cloud storage. The warehouse holds the JDBC session and runs the SQL.
In plain English
A SQL warehouse publishes a JDBC/ODBC endpoint — Server hostname plus HTTP path. Power BI, Tableau, and dbt open a session there, send SQL, and read rows back over the same connection. Photon executes it; the warehouse manages the query queue.
Without this step
Point BI at an all-purpose cluster instead and you inherit notebook semantics: no admission control, no per-query queue, and the endpoint dies the moment someone clicks Terminate.
One endpoint. Three very different machines can sit behind it — which starts in seconds?
