DuckDB
DuckDB is an open source, column-oriented SQL database engine built for analytical workloads. It runs in-process inside other applications, the way SQLite does for transactional work, and delivers fast aggregations over large datasets with zero configuration.
SQLite for analytics
DuckDB was created by the same research group at CWI that built MonetDB, one of the original columnar databases. It uses vectorized query execution, meaning it processes data in batches that fit modern CPU caches and SIMD instructions, rather than row by row. That makes it dramatically faster than row-oriented engines for analytical queries.
What makes DuckDB distinctive is that it needs no server process. You embed it directly into your application. It reads and writes Parquet files natively, supports standard SQL including window functions and CTEs, and can query data in place on disk, in S3, or over HTTPS without importing it first.
DuckDB has become the engine of choice for a growing number of analytical tools and databases that need a fast, embeddable SQL layer without the operational burden of a standalone database server.
How Arc handles DuckDB
Arc uses DuckDB as its SQL query engine. When you run a query against Arc, DuckDB executes it using vectorized columnar processing over your Parquet data. You get full analytical SQL with no separate database server to run or manage.
Related terms
Arc is a SQL-native time-series database. Open Parquet on storage you own, single Go binary, production-ready in 30 seconds.