Trino

Trino is developed in Java and designed for high-performance, interactive analytics. It supports:

• Full ANSI SQL compliance.

• A wide range of data sources through connectors, including:

  • Hive

  • Cassandra

  • Kafka

  • MongoDB

  • MySQL

  • PostgreSQL

  • SQL Server

  • Redis

  • Redshift

  • Local files

• Advanced data structures, including:

  • Arrays and maps

  • JSON data

  • GIS data

  • Color data

• Extensibility through multiple mechanisms, including:

  • Extension data connectors

  • Custom data types

  • Custom SQL functions

• Pipeline processing that streams query results to clients in real time.

• Monitoring through multiple interfaces:

  • A web UI that visualizes query execution progress.

  • The JMX protocol.

The following figure shows the system components of Trino.

A Trino cluster uses a coordinator/worker architecture. The coordinator accepts SQL queries from clients, parses and plans them, and distributes the resulting tasks to worker nodes. The coordinator handles:

• Accepting user queries, parsing them, generating execution plans, and distributing tasks to worker nodes.

• Monitoring worker node status. Each worker node maintains a heartbeat connection with the coordinator to report its status.

• Maintaining metastore data.

Trino is built for Online Analytical Processing (OLAP) workloads — data analysis, large-scale aggregations, and reporting. It is suitable for:

• Extract, transform, and load (ETL)

• Ad hoc queries

• Analysis of large-scale structured or semi-structured data

• Aggregation and report analysis of large-scale multidimensional data

Trino on E-MapReduce (EMR) offers the following advantages over open source Trino:

• Ready-to-use clusters. Spin up a Trino cluster with hundreds of nodes in minutes, with no setup required.

• Simple elastic scaling.

• Seamless integration with the EMR software stack and support for data stored in Object Storage Service (OSS).

• Fully managed operations with no O&M required.

Data model

Trino organizes data in a three-layer hierarchy: catalogs, schemas, and tables.

• Catalog

  A catalog maps to an external data source through a connector and contains one or more schemas. A single query can span multiple catalogs.

• Schema

  A schema is equivalent to a database instance and contains multiple tables.

• Table

  A table functions the same as in a standard relational database.

Connector

A connector adapts Trino to a specific data source, similar to a database driver. Each catalog is bound to one connector type, configured in the catalog's properties file.

Trino provides a standard Service Provider Interface (SPI) that you can implement to build custom connectors for any data source. Trino also ships with several built-in connectors grouped by data source type:

• Data lakes and lakehouses: Hive and other Hadoop-compatible sources

• Relational databases: MySQL, PostgreSQL, SQL Server

• Other systems: Cassandra, Kafka, MongoDB, Redis, Redshift, and local files