Analytical Java with DuckDB
| Event | J-Fall 2026 |
| Speaker | Geertjan Wielenga |
| Slide deck | Download |
Abstract
Your Java service just needs to answer a few analytical questions: top products this month, p95 latency per region, a quick join between a CSV export and a Parquet dump from the data team. The usual answers are to hammer your transactional database with GROUP BYs it was never designed for, ship the data off to a warehouse and wait, or write the aggregation by hand in streams. But did you know that there’s another option that runs inside your JVM…
DuckDB is a free open source in-process analytical database: columnar, vectorized, zero infrastructure, and available to Java through a plain JDBC driver.
In this session, we’ll use it to query Parquet, CSV, and JSON files directly, run analytical SQL over in-memory data, and push results straight into Java records and Apache Arrow.
Along the way, we’ll look at where it beats a traditional RDBMS, where it doesn’t, and how it fits into your services, batch jobs, tests, and CLIs… without any new operational setup or infrastructure.