DuckDB: Embedded Analytics with Parallel/Vector/Columnar Performance

Speakers:  Jim Tommaney

Jim Tommaney - Databricks - Percona Live 2021 - Altinity Community Track

DuckDB is a project coming out of CWI in the Netherlands that combines vector, columnar, and parallel capabilities.


  • Great performance for analytic queries
  • Fast batch load
  • Near linear scaling
  • Near-zero administration

DuckDB is a SQLite replacement for analytics.

In this session you will learn about:

  • What is different about Embedded/Serverless data engines?
  • Current performance and features of DuckDB
  • Best practices using DuckDB




Jim Tommaney

Databricks, Staff Performance Engineer

Currently performance tuning at Databricks. Previously architect of InniDB (now available as MariaDB Columnstore)

Hands-on experience delivering solutions across verticals including Telco, Mobile Marketing, CRM, Genomics, fraud analytics and more.

See all talks by Jim Tommaney »

✎ Edit this page on GitHub