Definition
Embedded OLAP databases run inside the application process (no separate server daemon), optimized for analytical SQL over columnar formats like Parquet—common in data engineering notebooks, ETL scripts, and edge analytics.
Key Points
- duckdb is the leading open-source embedded OLAP engine in 2026; v2.0 adds async object-storage I/O and faster recursive CTEs (2026-10-10-duckdb-2-0-performance-improvements)
- Cloud vendors (motherduck) extend the same engine with remote storage and compute separation
- Complements warehouse databases (Snowflake/BigQuery) for local iteration and pipeline unit tests