Confluent

Multi-Region Clusters with Confluent Platform 5.4

Sep 16, 2019David Arthur
Multi-Region Clusters with Confluent Platform 5.4
SummaryAI, from the source post

Introduces Multi-Region Clusters in Confluent Platform 5.4, built into Confluent Server. Details three key features: Follower Fetching (KIP-392) for consumers to read from the closest replica to reduce cross-WAN traffic and latency; Observers as asynchronous replicas that replicate data but do not join the ISR or become leaders, improving durability without impacting write throughput and preventing ISR thrashing; and a new JSON-based Replica Placement strategy for specifying replica counts and constraints across different 'racks' (locations). Also discusses multi-region ZooKeeper deployment considerations and KIP-500.

More from the Multi-Geo Replication trail