What it is
SQL Database in Fabric is the same Azure SQL engine you know — full T-SQL, row-store + columnstore, indexes, stored procedures, triggers — but provisioned and billed inside Fabric. The signature feature: every table is automatically mirrored to OneLake, so analytics workloads can read your transactional data with seconds-of-lag.
vs Warehouse
- SQL Database — OLTP. Frequent reads + writes, low-latency transactions, ACID. Many small queries.
- Warehouse — OLAP. Few big queries, columnar scans, append-mostly. Analytics-shaped data.
Both expose T-SQL surface; both store data in OneLake. They serve different workloads.
Auto-mirroring to OneLake
When you create a SQL Database, Fabric automatically provisions a mirrored Delta-Parquet copy in OneLake. Lakehouses, Warehouses, Power BI Direct Lake, and notebooks can all read it. The mirror is read-only, transactional, and auto-incremental.
When to use it
- Building a custom application that needs an OLTP store inside your Fabric tenant.
- You want analytics on transactional data without setting up Mirroring separately.
- You prefer Microsoft-native OLTP over Cosmos DB or external Azure SQL.
Best practices
- Use the auto-mirror — don't add a second Mirroring item over the same DB.
- Index for the OLTP pattern. The analytics path uses the OneLake mirror, not the source indexes.
- Treat the OneLake mirror as bronze. Promote into Silver/Gold lakehouse tables for serving.