Distributed databases store data across multiple servers managed as a single logical unit, enabling horizontal scaling and high availability. Two primary architectures are covered: primary/secondary (master/slave), where one node handles all writes and secondaries serve as read replicas, and shared-nothing (masterless), where data is partitioned across all nodes for linear write and read scalability. The post explains sharding, replica promotion for fault tolerance, and why relational databases struggle with distribution. Couchbase's approach using 1024 vBuckets with a client-side vBucket map for direct node routing is highlighted as an example of automatic shared-nothing distribution.
Table of contents
Primary/Secondary vs. Shared-Nothing ArchitectureHorizontal ScalingDistributed Database Examples638 Impressions