Can Rust’s Limbo Outpace SQLite? Inside Turso’s Async Database Rewrite

Turso, founded by Glauber Costa and Pekka Enberg in 2021, originally forked SQLite to create libSQL, citing the need for a modernized, open‑source version. Realizing that a simple fork could not deliver features like asynchronous I/O, Enberg launched a personal project to rewrite SQLite in Rust, which became the official Turso Limbo project.

Limbo (the project codename) strives for 100% compatibility with SQLite while providing full memory safety and a modern architecture. It is released under the MIT license on GitHub.

The project incorporates deterministic simulation testing (DST) to match or exceed SQLite’s reliability, leveraging Antithesis’s tooling for fault‑injection testing. Limbo also includes its own DST framework.

On Linux, Limbo uses the kernel’s io_uring system call—originally developed by a Meta engineer—to achieve high‑performance asynchronous I/O. The new engine extends sqlite3_step to return immediately when data is not yet ready.

Performance testing shows Limbo can be faster than SQLite on large data sets, though results vary. The developers emphasize that the goal is not to claim universal speed superiority but to demonstrate solid early‑stage performance.

Limbo can be compiled to WebAssembly (WASM), a feature highlighted by Turso’s co‑founders as a “next‑step for SQLite.”

Building Limbo on macOS requires the latest Rust compiler to resolve compilation errors.

Community feedback includes skepticism about rewriting SQLite, noting that the original’s code quality and extensive test suite make it a difficult target for replacement. However, proponents argue the rewrite enables features and safety guarantees that are hard to achieve in C/C++.