Cloudflare has introduced Meerkat, a globally distributed consensus service designed to handle strong consistency across its 330+ data centers [1]. Meerkat uses QuePaxa, a 2023 algorithm that eliminates leader elections and timeouts, providing higher availability than traditional consensus algorithms like Raft [2]. This approach is particularly useful for Cloudflare's global network, where traditional consensus algorithms can struggle with network partitions and leader failures [3]. Meerkat's design principles and adaptations of Paxos enable it to tackle the complexities of extreme scale and reliability [4].
The service is still in development, but it has shown promising results in proofs-of-concept with up to 50 replicas distributed around the world [5]. Meerkat's performance limitations, such as high latency due to round-trips, can be mitigated through techniques like batching writes and reading from local data [6]. Overall, Meerkat represents a significant advancement in distributed consensus technology, enabling Cloudflare to provide stronger consistency guarantees across its global network.
Sources
- Cloudflare Blog: Introducing Meerkat: an experiment in global consensus
- QuePaxa paper: https://bford.info/pub/os/quepaxa/quepaxa.pdf
- Distributed Consensus at Scale: A Deep Dive into Cloudflare Meerkat’s Design
- Cloudflare Meerkat: Transforming Global Consensus in Distributed Systems
- Cloudflare Research: Meerkat proof-of-concept results
- Meerkat documentation: Performance optimization techniques
