The scale is growing fast. Cloudflare reports a 15x increase in live non-deleted D1 databases over the past year and 65x over two years, with daily active databases up 18x year on year, as of 8 October 2026. A database counts as active if it ran at least one query that day.

The central engineering challenge was scaling the control plane, the machinery that manages database lifecycle and routing, without slowing every query by pushing it through a centralised bottleneck. Cloudflare's answer keeps the hot path, meaning all query requests, free of single-region components. Each D1 database is backed by a SQLite Durable Object, and Workers KV, a globally distributed key-value store, caches read-heavy routing metadata at the edge so cache misses only rarely hit central databases.

Management operations such as creating, deleting and listing databases do use a central Postgres cluster, accessed via Hyperdrive for connection pooling and caching, with read replicas handling cached reads. Cloudflare also stresses designing for failure and idempotency, using eventual consistency between components that do not need strong consistency, and load-testing against targets well above expected growth. It notes the patterns are applicable to anyone building control planes on its Developer Platform.