Multi-Geo Replication
Multi-Region Clusters with Confluent Platform 5.4

Multi-Region Clusters with Confluent Platform 5.4

9/16/2019 · David Arthur

What this post added

Introduces Multi-Region Clusters in Confluent Platform 5.4, enabling per-region data replication. Details Follower Fetching (KIP-392) allowing consumers to fetch from the closest replica to reduce cross-WAN traffic and latency. Explains Observers as asynchronous replicas that don't impact ISR or leadership, improving durability without affecting write throughput or causing ISR thrashing. Introduces a JSON-based Replica Placement strategy for specifying replica counts and constraints across regions, complementing Observers for disaster recovery. Discusses multi-region ZooKeeper deployment challenges and KIP-500.

Read the original post ↗