Databricks' now-generally-available connector queries Workday HR and finance data in Workday's own storage

Lakehouse Federation has two modes, and they put the work in different places. Query federation sends SQL to the external database and runs it on that system's compute ^[5]. Catalog federation, the mode this connector uses, takes table metadata from Workday's Iceberg REST catalog and leaves the scans and joins to Databricks ^[6].

The zero-copy label describes the Databricks side. The connector reads shared Iceberg tables that already sit in cloud storage Workday manages ^[4][6]. So there is one physical copy, and Workday keeps it. I think that is the right design for people data. With Workday holding the only copy, a data team has no second set of salary records to secure or reconcile ^[8].

Setup runs through two administrators, one on each platform ^[13]. Once the Workday catalog appears in Unity Catalog, permissions apply at catalog, schema and table level on every federated query ^[11]. Jonathan Rogers, chief information officer at FactSet, said: "The Databricks Workday Data Connect federation connector gives us a governed, zero-copy way to discover and catalog Workday data in Unity Catalog." ^[12]

Whether the extract pipelines go away depends on what they feed. Databricks describes three shapes for Workday data: a durable copy for historical reporting, an on-demand lookup, and a governed live view for analytics ^[10]. Federation is the live view ^[10]. For a durable copy with SCD2 history tracking, Databricks sends customers to Lakeflow Connect for Workday Reports and Workday HCM, a separate ingestion product ^[9]. A pipeline that lands current-state tables for dashboards and ad hoc SQL can be retired. A pipeline that keeps history, say for a headcount trend or a point-in-time pay audit, still has a job ^[9].

Workday Data Connect itself is generally available ^[2], while the Databricks connector that reads it is in Beta ^[1]. The announcement does not give a general-availability date for the connector, support terms, or read latency against Workday's storage. In my context, I would point notebooks and Genie spaces at the federated catalog first ^[14] and leave existing ingestion running for anything that feeds the financial close. Running both paths for a few reporting cycles and comparing row counts would show whether the live view matches the extracts. If the connector changes during Beta, the fallback is the Lakeflow Connect ingestion path Databricks already documents for history ^[9].