DuckDB is faster than SQLite on an analytical type of query. That's expected because DuckDB is column-based and designed exactly for this type of queries.
The analysis would be more telling if it measured insert performance as well.
Kind of. The dataset is small and he probably need to create the database at most once per day and have to read a lot more for his analysis/exploration.
Comments
DuckDB is faster than SQLite on an analytical type of query. That's expected because DuckDB is column-based and designed exactly for this type of queries.
The analysis would be more telling if it measured insert performance as well.
Kind of. The dataset is small and he probably need to create the database at most once per day and have to read a lot more for his analysis/exploration.
Columnar databases are a good fit here.
Or reading all fields of a very clustered 5% of rows, to give an even greater advantage to the row-oriented DBMS.
Mature users need to evaluate how their workloads match the benchmarks, it isn't a matter of DuckDB or SQLite "winning".