# Starburst

> Commercial Trino, self-managed or fully hosted

Starburst is the commercial platform built on Trino, sold as Starburst Enterprise for self-managed clusters and Starburst Galaxy as a fully managed service, adding governance, performance work, and a supported Iceberg lakehouse on top of the open-source engine.

Website: https://www.starburst.io

## Use it when

- You want Trino with someone accountable for it: the engine's creators work here, and Enterprise is a supported distribution.
- Governance is the gap plain Trino leaves open: built-in RBAC, row filters, and column masks managed in a UI.
- You need connectors Trino lacks (Snowflake, Teradata, SAP HANA, Salesforce, Db2, Splunk) among more than 60.
- You want a managed lakehouse quickly: Galaxy runs on AWS, Google Cloud, and Azure with a standing free plan, and Icehouse makes the Iceberg decisions for you.

## Think twice when

- Plain Trino already serves you and budget is the question; the open-source engine speaks the same dialect and covers most of the capability.
- You want to avoid vendor coupling on top of an open engine; the governance and acceleration layers are Starburst-only.
- Self-managed Enterprise still means operating Kubernetes clusters; only Galaxy removes the ops.

## How it runs

Enterprise deploys self-managed on Kubernetes, on-premises or in any cloud; Galaxy is fully managed and billed per worker-hour. Both run Trino's coordinator-plus-workers architecture with fault-tolerant execution spooling exchanges to object storage. Clients connect over JDBC, ODBC, and the Trino CLI.

## Details

### General

| Attribute | Value |
| --- | --- |
| License | Commercial |
| Deployment | Self-hosted, Managed |
| Workload | Interactive, Batch |
| Operational complexity | Medium |
| Pricing | Free, Subscription |
| Documentation | https://docs.starburst.io |

### Query Engines

| Attribute | Value |
| --- | --- |
| Engine shape | Distributed cluster; Galaxy is fully managed |
| Non-lake sources | Over 60 connectors: Trino's, improved versions of them, and Starburst-only ones for Snowflake, Teradata, Salesforce, SAP HANA, Netezza, Db2, Neo4j and Splunk |
| Table formats | Hive formats, Iceberg, Delta and Hudi through improved connectors; Icehouse is the opinionated Iceberg build |
| Catalog required | A Hive Metastore or Glue for the lake connectors; the cache service wants a relational database of its own |
| SQL dialect | Trino SQL: the same dialect as Trino, and close enough to Athena's to move between them |
| Client access | JDBC and ODBC drivers and the Trino CLI, plus the Galaxy web UI |
| Concurrent users | Resource groups queue and cap concurrency by user, source or query type, Trino's mechanism, sold with the management around it |
| Acceleration | The cache service automates materialized views and table scan redirection; Warp Speed still indexes unpartitioned columns, but its properties are deprecated and its caching moved to the file system cache in 479-e |
| Failure recovery | Trino's fault-tolerant execution: QUERY or TASK retry, with an exchange manager spooling to S3, Azure Blob, GCS, Alluxio or HDFS |
| Access control | Built-in role-based access control with masks and row filters in the web UI; Ranger, Immuta and Lake Formation integrate alongside |

## Capabilities

- [Query Engines](https://matca.io/capabilities/query-engine)

---

Source: https://matca.io/starburst
Last updated: 2026-08-31T11:40:46.794Z