DuckDB 2.0's quack server lets ten writers share one file for under 2ms a query

7.8 relevance

DuckDB 2.0's concurrent write server is a significant technical advancement with immediate practical use for data engineers.

DevTools dev.to

Answer the board on AI readiness

Arrive with evidence, not opinion. See how your foundation benchmarks against your peers, get a clear read, and leave with a ranked place to start.

Why It Matters

For platform engineers building multi-process data pipelines on DuckDB, quack eliminates the single-writer bottleneck without changing the storage layer, but the throughput ceiling at 5 concurrent writers means you still need careful client pooling for write-heavy workloads.

Key Takeaways

Benchmark quack with your actual write concurrency pattern—peak throughput is at 5 clients, not 10 or 20—and consider connection pooling to stay under the overhead cliff.

DuckDB 2.0's quack server lets ten writers share one file for under 2ms a query

Summary

DuckDB 2.0's quack extension turns a running DuckDB process into a network server, enabling up to ten concurrent writers to share a single file without file-lock conflicts. Benchmarks show throughput peaks at 5 concurrent clients (735 inserts/s) then degrades, with 20 clients matching single-client serial performance (~350/s). Quack does not bypass DuckDB's internal write serialization—it provides a single process gateway that multiplexes writes through the one allowed lock holder.

Author

Alex Georgiev