Cloudflare has developed a new distributed consensus service called Meerkat, utilizing the QuePaxa algorithm to enable all replicas to write data simultaneously. This approach addresses challenges faced by traditional consensus algorithms like Raft, particularly in variable network environments. Meerkat is currently in development and is intended for internal use to manage control plane state.
Cloudflare's internal services often require consistent control-plane state management across more than 330 global data centers. Traditional consensus algorithms struggle under unpredictable Internet conditions, leading to potential data inconsistencies and systemic downtime.
Common algorithms like Raft face limitations in wide-area networks because they depend on designated leaders and timeout configurations. If a leader fails, the system can become inoperable until a new one is elected, which can exacerbate delays in data synchronization.
In response to these challenges, Cloudflare's Research team has developed Meerkat, based on the QuePaxa algorithm. This new approach allows any replica to write data at any time, thereby ensuring that system progress is never stalled by leader failures or timeouts.
Meerkat is still in the experimental phase and designed primarily for internal operations, targeting small pieces of control plane state management. It is poised to be the first large-scale deployment of the QuePaxa algorithm in an industrial setting.
✨ This summary was generated by AI from the outlets' reporting listed below. It is not independently verified and may contain errors — check the original sources. How BrevFeed works →
Cloudflare has developed a new distributed consensus service called Meerkat, utilizing the QuePaxa algorithm to enable all replicas to write data simultaneously. This approach addresses challenges faced by traditional consensus algorithms like Raft, particularly in variable network environments. Meerkat is currently in development and is intended for internal use to manage control plane state.