Presto

Overview

Presto is an open-source distributed SQL query engine originally created at Facebook, where development began in 2012 and the project was open-sourced in 2013. It was built to run fast, interactive analytical queries over very large datasets, and to query data where it lives rather than requiring it to be loaded into a dedicated warehouse first. Today this original lineage, often called PrestoDB, is governed by the Presto Foundation, hosted at the Linux Foundation.

Architecture and connectors

Presto uses a coordinator-and-worker architecture: the coordinator parses, plans, and schedules queries, while workers execute tasks in parallel and stream data through memory rather than writing intermediate results to disk. A connector model lets Presto query many systems, including Hadoop/HDFS, object storage, and relational databases, and join across them in a single query.

Copy code
-- Presto: aggregate over data accessed via a connector
SELECT region, COUNT(*) AS orders, SUM(amount) AS revenue
FROM hive.sales.orders
GROUP BY region
ORDER BY revenue DESC;

Presto, Trino, and DuckDB

In 2019 several original creators of Presto forked the project as PrestoSQL, which was renamed Trino in December 2020; the two projects share ancestry but are now developed separately. Both Presto and Trino are distributed engines aimed at scaling across a cluster and federating queries over many sources. DuckDB takes a different approach: it is a single-node, in-process OLAP engine, with MotherDuck as its serverless cloud counterpart. For workloads that fit on one machine or for querying data lake files directly, DuckDB avoids the operational overhead of a coordinator/worker cluster:

Copy code
-- DuckDB: single-node query over lake files
SELECT region, SUM(amount) AS revenue
FROM read_parquet('s3://sales/orders/*.parquet')
GROUP BY ALL
ORDER BY revenue DESC;