# Cloudflare scales D1 SQLite database to millions of instances

> **Key takeaway:** Cloudflare has re-architected the D1 control plane to support millions of serverless SQLite databases using edge-based routing and distributed metadata.

**Published:** 2026-10-13T16:20:18+00:00 
**Source:** Cloudflare Technical Blog 
**Category:** software-engineering 
**Canonical URL:** https://fosswire.org/news/cloudflare-scales-d1-sqlite-database-to-millions-of-instance.html 

## Summary
Cloudflare has implemented architectural changes to its D1 serverless SQLite platform to handle massive growth, including a 65x increase in live databases over two years. The update focuses on decoupling the control plane from the data plane to prevent centralized bottlenecks. By utilizing Workers KV for routing metadata and Hyperdrive for Postgres connection pooling, D1 can now support millions of individual databases for multi-tenant and AI-driven applications.

## D1 Architectural Improvements
- Each D1 database is backed by a single **SQLite Durable Object**.
- **Workers KV** is used as a globally distributed caching layer for critical routing metadata to minimize central database queries.
- **Hyperdrive** manages connections to central Postgres clusters for CRUD operations and efficient connection pooling.
- The **D1 Eyeball Worker** acts as a proxy for all requests requiring Durable Objects, handling observability and retries.
- The architecture separates operations into three planes: **Data plane** (edge-based queries), **Control plane** (API/management), and **Background asynchronous plane** (analytics/billing).

## Why it matters
This scaling allows developers to move away from large, monolithic database clusters toward a 'one database per user/tenant' model, which is increasingly required by AI-driven development platforms.