Starburst

Starburst extends Trino with enterprise features: data products, data access control, column masking, row-level security, cost-based optimizer improvements, and a managed cloud offering (Starburst Galaxy). The core capability is federated SQL: run a single query that joins data from S3/Parquet, Snowflake, a Postgres database, and a Kafka topic — no ETL required. For agents, this is powerful for cross-source analytical queries that would otherwise require complex pipeline orchestration. Starburst Galaxy (SaaS) simplifies cluster management; Starburst Enterprise supports self-hosted Kubernetes deployments. REST API matches the Trino REST protocol. JDBC preferred for broad BI tool compatibility. Confidence is docs-derived.

REST API follows Trino protocol at {host}:8080. POST /v1/statement executes a SQL query and returns a nextUri for polling results. GET {nextUri} continues fetching result pages. DELETE {queryId} cancels a running query. Column metadata in response header; data rows in JSON format. JDBC driver compatible with any Trino-compatible JDBC client. CLI available for interactive queries. Starburst Galaxy has a UI-based query editor plus the same REST/JDBC endpoint. Catalog registration (data source connectors) done via Galaxy admin UI or starburst-admin API.

Starburst Galaxy: OAuth2 with service account tokens (created via Galaxy Console) sent as Authorization: Bearer. Starburst Enterprise: LDAP, Kerberos, OAuth2, or password auth depending on configuration. Username/password sent via Authorization: Basic on the REST API or in JDBC connection string. TLS enforced. Galaxy role-based access: viewer, editor, admin at catalog/schema/table level. Column masking and row-level filtering for fine-grained security.

Trino/Starburst REST protocol: query results paginated via nextUri polling. Failed queries return error object with errorCode, message, and errorLocation (SQL position). Resource group limits prevent runaway queries. Query queue with configurable max queued time. Cluster auto-scaling in Galaxy. Self-hosted reliability depends on coordinator/worker configuration. Starburst Galaxy SLA: 99.9%. Query memory limits configurable per query/session. Remote source errors (e.g., Snowflake connector timeout) surface with connector-specific error codes.

docs.starburst.io provides Trino SQL reference, connector documentation (S3/Hive, Snowflake, BigQuery, MySQL, Postgres, Kafka, etc.), REST API reference, security configuration, Galaxy setup guide, and data product creation. Documentation is extensive and well-organized. Getting started requires a Starburst Galaxy account (free trial) or Starburst Enterprise license. JDBC driver download available. Community support via Trino.io (upstream project) and Starburst Community Slack.