Query Engines on the Lakehouse

A lakehouse separates storage from compute, so you pick the engine per workload. The same Iceberg 129 tables now meet Trino 483 403,499 (18 July 2026, still current) as the shared SQL service, Spark 4.1.3 129 as the heavy writer, and DuckDB 1.5.5 61,228 as the in-process client (1.5.6 of 28 September 2026 is a patch release). None owns the data.

Subsections

• Trino's Architecture
• Trino with Iceberg
• Querying with Trino
• Optimizer and Federation
• Spark as a Lakehouse Engine
• DuckDB Iceberg and Delta
• Choosing an Engine