Turso

Former virtualization engineers Glauber Costa and Pekka Enberg created Turso and its underlying libSQL engine to break SQLite out of its embedded boundary and turn it into a distributed database for modern microservices. Because upstream SQLite rejects external code contributions, they forked the codebase to add remote HTTP connections, automated replication, and multi-tenant architectures. Today, that lineage gives infrastructure teams a serverless database that delivers local in-process read latency while syncing writes back to a primary coordinator.

Why Turso instead of PlanetScale

PlanetScale organizes relational data around centralized Vitess clusters running MySQL, designed to maximize throughput for large services hosted in core cloud datacenters. While this architecture handles heavy transactional loads, applications deployed across global edge networks or regional worker nodes must cross public internet links to execute every query. That network traversal adds thirty to one hundred milliseconds of roundtrip latency per request, forcing engineering teams to introduce auxiliary caching tiers like Redis simply to keep edge responses responsive.

Turso approaches database performance by inverting the network model: instead of shuttling queries to a distant cluster, it pushes read replicas directly to application nodes or edge locations. Using embedded replicas, queries run against local disk or memory with sub-millisecond execution times, while write operations automatically route back to the primary database over an optimized protocol. Teams migrating from PlanetScale trade MySQL's horizontal sharding and Git-like branching for zero-cold-start SQLite databases, negligible read latency, and dramatically simpler operations when distributing data close to end users.

What it does well

• Embedded replicas place SQLite databases on local disk or inside memory within application containers, eliminating network hops on read queries.
• Multi-database architecture allows teams to provision hundreds or thousands of isolated, lightweight databases for multi-tenant SaaS applications at minimal resource cost.
• Native HTTP and WebSocket wire protocols enable direct database connections from serverless and edge runtimes without persistent TCP connection pool exhaustion.
• Full SQLite file-format and query compatibility ensures existing SQL queries, local dev environments, and standard SQLite tools work without rewriting database logic.