Technology

Trino

A distributed SQL query engine for federated analytics across heterogeneous data sources, with deep support for S3-backed data lakes and lakehouses.

11 connections 4 resources

Summary

What it is

A distributed SQL query engine for federated analytics across heterogeneous data sources, with deep support for S3-backed data lakes and lakehouses.

Where it fits

Trino is the multi-engine query layer for S3 lakehouses. It queries Iceberg, Delta, Hudi, and raw Parquet on S3 through connectors — and can join S3 data with operational databases in a single query.

Misconceptions / Traps

Trino is a query engine, not a storage engine. It reads from S3 but does not manage data. Writes go through table format commit protocols.
Trino requires a coordinator and workers — operational overhead is higher than DuckDB. Use DuckDB for single-user exploration; Trino for multi-user production queries.

Key Connections

depends_on Apache Parquet — reads Parquet files from S3
used_by Lakehouse Architecture — a primary query engine for lakehouses
constrained_by Small Files Problem, Object Listing Performance — performance affected by S3 access patterns
Natural Language Querying augments Trino — LLMs generate SQL for Trino
scoped_to S3, Lakehouse

Definition

What it is

A distributed SQL query engine designed for federated analytics across heterogeneous data sources, with deep support for querying S3-backed data lakes and lakehouses.

Why it exists

Organizations store data across many systems. Trino provides a single SQL interface to query data wherever it lives — including directly on S3 via Parquet, ORC, Iceberg, Delta, and Hudi connectors — without moving data.

Primary use cases

Federated SQL across S3-backed sources, interactive lakehouse queries, cross-source joins between S3 data and operational databases.

Recent developments

Latest signals

Trino 478 and 479 ship community broadcast cadence. Per the Trino Community Broadcast episode list, Trino 478 and 479 continue the project's high-cadence release pattern with topics covering virtual view hierarchies (with Rob Dickinson), AI agents for query development, Trino Query UI updates, and Trino Gateway 16 → 18 progression. The community-broadcast cadence is itself a competitive signal: Trino maintains the analyst-engagement velocity that closed-source data-warehouse vendors struggle to match.
Trino Gateway 18 — in-memory caching + Java 25. Per the Trino Gateway release notes, Gateway 18 adds in-memory caching of backend metadata, query-history deactivation, and UI timezone selection. Gateway 17 (the predecessor) requires Java 25 and ships on the UBI10 micro base image with JMX metrics enabled. For organizations running Trino as a federated query layer over S3-backed lakehouses, the Gateway tier is now first-class infrastructure with its own release cadence rather than an afterthought.