Microsoft Fabric Lakehouse SQL

BenchBox supports Microsoft Fabric Lakehouse SQL for read-only query benchmarking.

Key Constraint

Fabric Lakehouse SQL endpoint is read-only:

• DDL is rejected (CREATE, ALTER, DROP)
• DML is rejected (INSERT, UPDATE, DELETE, MERGE)
• Query execution (SELECT) is supported

Use fabric-spark for generate,load phases, then run query phases on fabric-lakehouse.

Installation

uv add benchbox --extra fabric

Typical Workflow

benchbox run --platform fabric-spark --benchmark tpch --scale 1 --phases generate,load --non-interactive

benchbox run --platform fabric-lakehouse --benchmark tpch --scale 1 --phases power,throughput --non-interactive

Step 1 loads data through Spark. Step 2 runs the query phases through the Lakehouse SQL endpoint.

Adapter Configuration

• workspace: Fabric workspace name or GUID
• lakehouse: Lakehouse database name
• server: Optional explicit SQL endpoint (<workspace>.sql.azuresynapse.net)
• auth_method: default_credential, service_principal, or interactive

Notes

• Dialect: tsql
• Authentication: Entra ID only
• Data loading and schema creation must be done through Spark