Petabyte-scale.

Sub-second response

Starburst's analytics engine is powered by Trino: the open source query engine we helped create. Query hundreds of billions of rows across 50+ sources in a single SQL statement. No copying. No duplicates.
Start Free →
Starburst Analytics Engine
faster than Spark on typical analytic workloads
3.2×
native connectors: Amazon S3, Snowflake, Postgres, and more
50+
query across any combination of sources, no ETL
1 SQL
horizontal scale: add workers in seconds
∞

How it works

Built for data federation first. Not bolted on after

Most engines started as a single data store and added connectors later. Starburst's engine was built from day one to federate: query anything, from anywhere, fast. That's the Trino heritage.

Cross-source data federation

One SQL query. Every data source. No ETL

Write a single SQL statement that joins a Snowflake table, an Amazon S3 Iceberg table, a live Postgres database, and a Kafka stream. The engine pushes predicates down to each source and assembles results at query time: no copies, no pipelines, no waiting.

  • ANSI-standard SQL works across all sources
  • Predicate pushdown to each native source
  • Cross-source JOINs with cost-based join ordering
  • 50+ certified connectors: JDBC, REST, and native
Cross-source data federation — one query across every source

Query optimizer

The engine that thinks before it runs

Starburst's query optimizer analyzes table statistics, partition layouts, and cluster load before executing a query: choosing the most efficient join order, applying dynamic filters, and distributing work across workers for maximum parallelism.

  • Automatic join reordering based on cardinality estimates
  • Dynamic partition pruning: skip irrelevant data early
  • Materialized view selection: serve from pre-computed results when fresher than threshold
  • Adaptive query execution: replan mid-flight on skewed data
Query optimizer — cost-based execution stages

Elastic scale

Add workers in seconds. Pay for what you use

Starburst's engine scales horizontally: add compute workers on demand during peak workloads, shrink during off hours, and isolate workloads with dedicated clusters per team or use case.

  • Auto-scaling in Starburst Galaxy: near-zero manual tuning
  • Dedicated clusters per team for workload isolation
  • Spot/preemptible instance support for cost optimization
  • Serverless query mode for bursty, infrequent workloads

Cluster utilization — auto-scale event

auto-scale +12 workers

Peak query load at 6 a.m. Cluster scaled from 8 to 20 workers automatically, no intervention required.

See what changes when intelligence reaches all your data.

Start free in minutes, or talk with a Starburst expert.

Start Free →Talk to an Expert →

No credit card required.