Skip to content

v0.1.0

Latest
Latest

Choose a tag to compare

View all tags
@matteobovetti matteobovetti released this 31 Jul 16:22
· 2 commits to main since this release
v0.1.0
7e1ea80

0.1.0 - 2026-07-31

First public release of VaireDB, a cloud-native, distributed SQL database that
combines PostgreSQL wire compatibility with DuckDB's columnar vectorized
execution engine for high-throughput analytical workloads across a horizontally
scalable cluster.

Important: v0.1 is an early, under-active-development release. Breaking changes
may occur and major features may still be added. It is not yet ready for
production use — contributions toward the production-ready v1.0 are welcome.

Highlights

  • PostgreSQL wire protocol (v3) — connect from any standard PostgreSQL client
    (psql, JDBC, any PG driver) on port 5432.
  • PostgreSQL-compatible SQL — parsed via DataFusion with automatic dialect
    translation to DuckDB, including pg_catalog emulation for client compatibility.
  • Horizontal scalability — data is hash-sharded across core nodes, each running
    an embedded DuckDB instance tuned for OLAP.
  • Analytical performance — DuckDB's columnar, vectorized engine, with distributed
    reads planned and executed through an embedded Ballista scheduler.
  • Fault tolerance — configurable replication factor with quorum-based writes and a
    background retry/backoff loop that tails missed writes to lagging replicas.
  • Column pseudonymization (compliance) — declared columns are HMAC-SHA256 hashed
    in the coordinator so plaintext never reaches storage; secret keys stay in the
    coordinator catalog.
  • Operational simplicity — self-contained Rust binaries (Rust 1.95, 2024 edition)
    configured with YAML.

Added

  • Coordinator and core node binaries with gRPC-based cluster coordination
    (registration, heartbeat-based failure detection).
  • SQL DDL/DML/SELECT support with PostgreSQL → DuckDB dialect rewriting.
  • Hash sharding with configurable shards, replication_factor, and shard_by.
  • Quorum writes with a background replica-catch-up retry loop.
  • Column pseudonymization via anonymized_columns in CREATE TABLE (HMAC-SHA256
    computed in the coordinator).
  • Persistent metadata catalog (redb) exposed as queryable virtual tables.
  • Embedded Ballista scheduler for distributed read planning with shard affinity.
  • Full documentation set (architecture specs, compliance feature design) and a
    published docs site.
  • End-to-end test suite (make e2e).

Architecture

A coordinator/worker topology:

  • Coordinator node — client connections, SQL parsing, query routing and
    distributed planning, metadata catalog, write dispatch, replication, and
    anonymization.
  • Core nodes — embedded DuckDB storage, shard-local query execution, and write
    replication.

Known limitations (non-goals for v0.1)

  • Single coordinator (no HA — known SPOF).
  • No online resharding; the sharding strategy is fixed at table creation/load time.
  • Manual failover and recovery.
  • No built-in authentication, authorization, or encryption.
  • No metrics/tracing endpoints.
  • Only the 1-coordinator + N-core-nodes topology is supported.

See Design Goals for the full list of goals and non-goals.

Roadmap to v0.2

  • Compliance data take-out and deletion (DAG-based).
  • Extended DuckDB SQL support.
  • VaireDB CLI with bulk-import SQL command and psql client.
  • Micro-benchmarks and distributed performance tests.

See the full Roadmap.

Assets 2
Loading