Neki

Neki is sharded Postgres by the team behind Vitess. It is architected from first principles to bring Vitess-level scale and reliability to PostgreSQL workloads.

Why Neki

What is Neki?

Vitess transformed MySQL sharding from an operational nightmare into an accessible, proven technology. It powers some of the largest sites on the Internet: Slack, GitHub, Square, and hundreds more by making it possible to scale MySQL horizontally across thousands of servers.

Neki brings this same power to PostgreSQL.

But Neki is not a fork of Vitess. Vitess succeeds by leveraging MySQL's strengths and engineering around its weaknesses. Postgres is a fundamentally different database with its own architecture, replication model, and operational characteristics. To achieve Vitess-level capabilities for Postgres, Neki is built from scratch — applying everything we've learned from years of operating sharded databases at massive scale.

Built by operators, for operators

The PlanetScale team runs some of the world's largest sharded Vitess clusters. We manage thousands of production workloads and have deep expertise in what it takes to keep distributed databases running reliably at scale.

Neki has this operational knowledge at its core. Every architectural decision is informed by real-world production experience:

• Failure handling: We know how sharded systems fail because we've lived through those failures and built systems to prevent and recover from them
• Performance optimization: We understand query routing, connection pooling, and distributed transaction patterns because we optimize them daily
• Operational simplicity: We've felt the pain of complex operational procedures, so Neki is designed to minimize toil and maximize automation
• Scaling patterns: We know which sharding strategies work at scale and which create operational headaches

Neki is the tool we wished we had when we started scaling Postgres.

Get started

Sign up for a PlanetScale account to use Neki today. Start with the Neki documentation to create a database, connect, and shard tables. If you already run PlanetScale Postgres, Coming from PlanetScale Postgres covers what changes.

A well-tuned single Postgres cluster can still be the right first step. Our Metal nodes can support extremely high throughput and provide performance that you don't get with network-attached storage solutions. We have NVMe drives that can hold up to 120 TB.

Questions? Contact us at neki@planetscale.com or reach out.

Related resources