Neon

Neon is a serverless open-source PostgreSQL database. It separates storage and compute to offer features like “Scale to Zero”, “Instant Branching”, and “Bottomless Storage”.

When to Use

• Cost Efficiency: “Scale to Zero” means you pay nothing when no one is using the app (Dev/Test envs).
• Developer Workflow: “Instant Branching” (Copy-on-write) lets you create a full clone of specific Prod data for testing in 1 second.
• Standard Postgres: It is Postgres. Not a fork with limitations.

Core Concepts

Separated Architecture

• Compute: Stateless Postgres nodes. Can boot in <500ms.
• Storage: Custom “Pageserver” layer purely for storage.

Data Branching

You can fork your database at a specific timestamp or LSN.

# Create a branch for testing feature X
neon branch create test_feature_x --from main

This is instant (COW) and cheap.

Best Practices (2025)

Do:

• Use Branching for CI/CD: Spin up an ephemeral DB branch for every Pull Request. Run tests, delete branch.
• Use Connection Pooling: Neon includes PgBouncer proxy. Use the pooled connection string for serverless functions.
• Autoscaling: Configure autoscaling limits to handle spikes without overpaying.

Don’t:

• Don’t worry about storage size: Neon storage scales automatically.

References

• Neon Documentation