ArnebDistributed SQL Query Engine

A Trino alternative built in Rust. Federated queries across heterogeneous data sources, served over the PostgreSQL wire protocol and Trino's client protocol.

Get Started

GitHub

Release notes

Faster and Leaner than Trino

On all 22 TPC-H queries, cell-identical results. SF10 2.3× faster warm at about a seventh of Trino's peak memory; SF30 3.5× faster cold at 0.22× the memory.

See the numbers

Arrow-Native

All intermediate data in Apache Arrow columnar format. No row-by-row processing.

PostgreSQL Compatible

Full Simple and Extended Query protocol. Works with psql, DBeaver, JDBC, and psycopg2 out of the box.

Trino Client Compatible

Trino's REST protocol, so the trino CLI, Trino JDBC, Superset, Metabase and dbt-trino connect unchanged.

Federated Queries

Query CSV and Parquet files, S3/GCS/Azure object stores, Hive Metastore catalogs (Parquet, ORC, partitioned and Iceberg tables) and PostgreSQL databases from a single SQL interface, and join across them in one statement.

Cross-source joins

Distributed Execution

Coordinator-worker architecture with Apache Arrow Flight RPC for high-throughput data exchange.

Secure and Governed

TLS and password authentication on both client protocols, a secret plus TLS between nodes, a query queue, per-query memory limits and statement timeouts.

Configure it

Observable and Fail-Safe

Prometheus metrics, a slow-query log, a Grafana dashboard and system.runtime.queries; a lost worker fails its queries with a clear error instead of hanging them.

Monitor it