[victorbjorklund]

DuckDB is a columnar db and focused on analysis queries. So it is more optimized for those workflows. Of course for many things sqlite is fine but if you have a really large db with complex queries duckdb should be faster.

[filleokus]

DuckDB also compresses the data. I've ingested ≈ 1 TB of apache-like access logs into a duckdb file that was small enough to load completely into ram on my local machine (like 10-15 GB). It was blazingly fast to query over.

(The large spark cluster was still ingesting the data when I was done with my analysis, only working on my local laptop)