Stratum is a columnar SQL engine built on the JVM exploring two ideas:
• SIMD execution using the Java Vector API
• copy-on-write branching for tables
It speaks the PostgreSQL wire protocol, so tools like psql, JDBC, and DBeaver work out of the box.
The engine is pure JVM (no JNI). In benchmarks on 10M rows it performs competitively with DuckDB on many analytical queries.
GitHub repo: https://github.com/replikativ/stratum
Benchmarks and methodology are described in the docs.
Happy to answer questions - feedback very welcome.