Open Data Infrastructure in practice: unlock your dbt projects with Apache Iceberg and mesh
Fivetran + dbt Labs demoed a real cross-platform dbt Mesh running on Apache Iceberg, no data copies required.
Overview
Most dbt projects face vendor lock-in from storage, SQL dialect, and query-engine constraints. Shows how dbt with Apache Iceberg, dbt v2's multi-dialect capabilities, and DuckDB innovations lets organizations operate across multiple platforms and engines for flexibility, collaboration, and cost savings.
What you'll take away
- 01
A new catalogs.yml v2 spec decouples dbt's Iceberg configuration from any one warehouse, letting a project define where its tables live independently of who writes to them.
- 02
The live demo proved the concept across three teams and two warehouses, Snowflake and Databricks, sharing one set of Iceberg tables through catalog federation, no data copying required.
- 03
Fivetran + dbt Labs’ new Lake Compute lets specific dbt models run on a cost-optimized engine, like DuckDB, than the rest of the project, configured with a simple adapter setting instead of a platform migration.
Session speakers
Jack Lowery
FivetranAnna Lee
dbt Labs