Facebook trapped in MySQL ‘fate worse than death’
According to database pioneer Michael Stonebraker, Facebook is operating a huge, complex MySQL implementation equivalent to “a fate worse than death,” and the only way out is “bite the bullet and rewrite everything.”
Context & Ripple Effects
Michael Stonebraker's verdict lands against a backdrop of years spent stretching one database far past its design point: Facebook has publicly walked through the software stack behind what was then the world's largest site, and GigaOM has already documented how the company wrings extra life out of commodity machines rather than buying its way out of bottlenecks (How Facebook Squeezes More From Its Machines).
The critique matters because it comes from the designer of Ingres and Postgres, aimed at the most-watched deployment in consumer web infrastructure — and because Facebook's growth story (Can Anything Stop The Facebook Juggernaut?) depends on a data layer that Stonebraker argues has become too patched and too complex to keep evolving.
First-order effects
- Facebook's infrastructure organization now faces a public fork: keep layering fixes onto a MySQL deployment Stonebraker calls unsalvageable, or take his advice to 'bite the bullet and rewrite everything' — a multi-year engineering commitment at production scale.
- The MySQL community and its commercial backers get their highest-profile stress test yet, with the largest social dataset on the web cited as evidence that patch-and-scale has hit its ceiling.
Second-order effects
- Startups pitching next-generation data stores gain a ready-made sales argument: if Facebook's approach fails by Stonebraker's measure, every buyer scaling past a single box has reason to weigh a rewrite over more sharding.
- Rival web platforms watching Facebook's predicament can position their own storage choices as the disciplined alternative, turning an internal architecture debate into competitive messaging.
Third-order effects
- If the largest operators conclude that general-purpose relational databases cannot follow their growth curves, database innovation migrates in-house — hyperscale sites funding custom engines and open-sourcing them rather than waiting on vendors.
- That shift would restructure the database market around operator-built systems, pressuring incumbent relational vendors to prove relevance at web scale or cede the fastest-growing workloads.
The trend: Web-scale operators are drifting from off-the-shelf relational databases toward self-built and open-sourced storage engines as growth outruns what patching a general-purpose system can deliver.