H2 Much Faster than DuckDB for small inserts / selected

[ryanhamilton]

Hi,

Today I've been running benchmarks and on small data sets H2 is inserting then selecting a subset 14x quicker than DuckDB.

I've got a data engine written in java that uses H2 as a backend. It allows performing selects against streaming data and other databases. So my use case is unique in that I create lots of small tables 100s of time per second, query them, then drop the table. I wanted to look at using DuckDB as the parquet querying is really useful to my users but H2 is so much faster I wouldn't change.

To be fair to DuckDB they are targeted at much larger data, are having to cross the java memory barrier and I wasn't using their custom appender.

Anyway just wanted to share the news and say Thanks.
👍

[grandinj]

Thanks!