ARCHITECTURE Signal 202
Introducing Meerkat: an experiment in global consensus
↗
Engineers operating distributed systems across wide-area networks regularly face availability problems when leader-based consensus algorithms like Raft lose their leader to crashes or network degradation—Cloudflare itself has suffered multiple incidents from this. QuePaxa's leaderless, timeout-free design could offer a more resilient foundation for globally distributed control-plane state if it proves viable at production scale.
Written by elseif from the cluster below · every claim links back to a sourceThe three things worth knowing
QuePaxa allows all replicas to perform writes at all times and never halts progress due to timeouts, unlike Raft which relies on a single leader and timeout-based elections that struggle over unpredictable WAN links.
Meerkat will initially manage small pieces of internal control-plane state such as leadership information for replicated databases, with a transactional key-value store and leasing system layered on top of its consensus log.
This represents the first industrial deployment of QuePaxa at global scale, though the service remains experimental and internal-only for the immediate future.
THE CLUSTER