330 data centers around the world, agreeing on the same thing at the same time - Cloudflare Research showed how: the Meerkat service, built on the QuePaxa algorithm, with about 10 times higher throughput than Raft under a bad network, per the company.
- Meerkat: Cloudflare's global consensus service, based on the algorithm QuePaxa.
- Keeps consistent state across more than 330 data centers; ~10x higher throughput than Raft under bad conditions.
- Per Cloudflare - the first industrial deployment of QuePaxa at global scale.
At first glance, a boring story. It comes down to the hardest problem in distributed systems: how do hundreds of machines around the world agree on what's true, when the network between them sometimes lies.
Here's the key. Raft was built for machines in one room, connected by a fast, predictable network. Stretch it across continents, and every delay spooks it. QuePaxa starts from somewhere else - it assumes the network is slow and moody, and works anyway.
For most of us this is infrastructure we'll never touch directly. I don't know if we'll feel QuePaxa directly in our own stack. But it's worth knowing: when the service on top holds up under a bad network, someone's deliberate decision sits behind that. It's not a given.