Trino was originally created as Presto at Facebook in 2012 by Martin Traverso, Dain Sundstrom, David Phillips, and Eric Hwang. The team left Facebook and rebranded the project to Trino in 2020 after a trademark dispute, while Facebook’s fork continues as PrestoDB.
The engine was built to solve a specific problem at Facebook: querying massive datasets interactively without waiting minutes or hours for batch jobs to complete. Presto let analysts run SQL queries across petabytes of data in Facebook’s data warehouse and get results in seconds to minutes.
Trino’s architecture is fundamentally different from data warehouses. It doesn’t store data — it’s a query engine that federates across data sources. A single Trino query can join tables from a data lake on S3, a PostgreSQL database, a MySQL instance, and an Elasticsearch cluster. This federation capability is Trino’s superpower.
The connector system is what makes federation work. Each connector knows how to read from a specific data source and push down filters and aggregations when possible. There are connectors for over 30 data sources including Hive, Iceberg, Delta Lake, Cassandra, MongoDB, Kafka, and standard relational databases.
Trino processes queries using a distributed pipeline of operators across worker nodes. It’s designed for high concurrency — many users running queries simultaneously — which makes it suitable as the query layer for BI tools and data exploration platforms.
Starburst, the commercial company founded by the Trino creators, provides a managed and enterprise version. Amazon Athena runs on a fork of Presto/Trino, making it one of the most widely deployed SQL engines in cloud environments.
For organizations with data spread across multiple systems that need a single query interface without consolidating everything into one warehouse, Trino is often the answer.
