Cloudflare picks leaderless QuePaxa consensus for Meerkat control plane
Original title: Choosing QuePaxa, an unusual consensus algorithm
Meerkat needs progress across continents when latency spikes, which leader-based Raft and Paxos do not guarantee.
Cloudflare says experimental Meerkat, its internal global control-plane consensus service, uses a new algorithm called QuePaxa instead of Raft, Multi-Paxos or Viewstamped Replication. Those leader-based systems slow or stop when the leader’s wide-area latency rises. Cloudflare cites a 2020 switch failure that made a consensus-backed dashboard about 80 times slower for six hours without data loss. The post explains destructive interference in Paxos and why QuePaxa avoids relying on a single leader.
- Meerkat replicas agree on an ordered log used for a key-value store.
- Comparable systems named are Consul, etcd and Apache ZooKeeper.
- A 2020 core switch failure left writes unavailable for six hours.
- QuePaxa is meant to keep progressing without a designated leader.
- Vendors
- Cloudflare
From the original article
14 min read at blog.cloudflare.com
Cloudflare is building an experimental consensus service using a never-before-deployed consensus algorithm, QuePaxa. In this post we explain how QuePaxa works and why we chose it.
By João Pedro Leite · blog.cloudflare.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.