multimaster

MultiMaster is a term used to describe a class of distributed-system architectures in which multiple nodes act as master replicas and can accept writes, rather than routing all updates through a single primary site. This active-active approach is designed to improve availability, reduce write latency, and provide regional write capability across data centers or geographic locations.

In a multi-master setup, data is replicated between nodes to keep the system synchronized. Replication can be synchronous, asynchronous, or a combination, depending on the system’s consistency requirements. Because writes may occur concurrently on different masters, conflicts can arise. Conflict handling and reconciliation are central challenges and are addressed through methods such as last-writer-wins, vector clocks and versioning, operational transformation, or the use of conflict-free replicated data types (CRDTs) to allow automatic conflict resolution.

Common implementations and ecosystems include databases and data stores that support active-active configurations, such as Galera Cluster for MySQL and MariaDB, CouchDB’s multi-master replication, and Cassandra’s multi-datacenter write paths with tunable consistency. Riak and other CRDT-based systems also embody multi-master principles. The choice of consistency model, latency requirements, and conflict-resolution strategy shapes deployment in e-commerce, SaaS, and globally distributed applications.

See also: active-active replication, multi-datacenter replication, CRDTs, conflict resolution, eventual consistency.