Replication
How 14 popular databases copy data across nodes for durability, availability, and read scaling.
PostgreSQL
Uses WAL-based streaming replication to maintain read replicas, replicating either the whole cluster (physical) or selected tables (logical). Async replication is the default, with synchronous replication available for stronger durability guarantees.
MySQL
Replicates via the binary log, historically in an async primary-replica topology. Semisynchronous replication and Group Replication add stronger guarantees, with the latter enabling multi-primary configurations.
SQLite
Has no built-in replication mechanism since it’s an embedded, single-file database. Multi-node durability relies on external tools like Litestream or application-level file copying.
MongoDB
Organizes nodes into replica sets with automatic, Raft-like primary election. Replication is asynchronous by default, but write concern settings (e.g., majority) let applications trade latency for durability.
Redis
Replicates asynchronously from a primary to one or more replicas. Redis Sentinel adds automated failover, while Redis Cluster combines sharding with per-shard replica sets for both scale and availability.
Cassandra
Peer-to-peer with no single primary; every node can accept writes. A configurable replication factor determines how many copies of each partition exist across nodes and datacenters, with consistency tuned per operation.
DynamoDB
Synchronously replicates data across multiple Availability Zones under the hood, transparent to the developer. Global Tables extend this with asynchronous multi-region replication for globally distributed applications.
CockroachDB
Replicates data ranges using the Raft consensus protocol, requiring a quorum of replicas to acknowledge writes. This makes replication synchronous by design, prioritizing strong consistency.
MariaDB
Inherits MySQL’s binary-log replication lineage, and additionally offers native Galera Cluster support for synchronous, multi-primary replication across nodes.
Oracle Database
Provides Data Guard for physical or logical standby replication aimed at disaster recovery, and GoldenGate for flexible, often multi-master, heterogeneous replication scenarios.
SQL Server
Always On Availability Groups support both synchronous and asynchronous replicas for HA and disaster recovery, alongside older transactional replication and log shipping options.
Elasticsearch
Each shard has a configurable number of replica shards that asynchronously mirror its data across the cluster, providing both fault tolerance and additional capacity for serving read/search traffic.
Neo4j
Causal clustering separates core servers, which replicate writes via Raft, from read replicas that asynchronously pull updates to scale out read throughput.
Firestore
Fully managed multi-region synchronous replication is built in when a database is configured in multi-region mode, entirely abstracted away from application code.