Skip to main content
Definition Website Computer and Technology

Database Replication

Subject: Computer and Technology noun /ˈdeɪ.təˌbeɪs ˌrɛp.lɪˈkeɪ.ʃən/ 1 definition

  1. Definition (noun)

    Entry #65 · Updated

    Database Replication is the automated process of copying and maintaining synchronized copies of data across multiple database servers, known as nodes or replicas, in real time or near real time. Rather than relying on static periodic backups or partitioning datasets into disjoint shards, replication propagates ongoing write operations—such as inserts, updates, and deletes—across nodes using mechanisms like transaction logs, write-ahead logs, or change streams. Its primary purpose is to eliminate single points of failure, enable automatic failover and high availability, reduce latency by serving reads closer to users, and scale read throughput while balancing consistency guarantees against replication lag.

    Common Forms

    • Single-Leader (Primary-Replica) Replication: Routes all write operations to one primary node, which streams change logs to one or more read-only replicas to scale read throughput and support failover.

    • Multi-Leader (Multi-Primary) Replication: Allows multiple designated nodes—often in separate geographic regions—to accept writes independently and exchange updates, improving regional write availability while requiring conflict resolution.

    • Leaderless Replication: Coordinates reads and writes directly across multiple replica nodes using quorum thresholds, avoiding a single write coordinator so clusters remain available during node failures.

    • Synchronous and Asynchronous Replication: Synchronous replication waits for replica acknowledgment before confirming a commit to prevent data loss, whereas asynchronous replication commits immediately on the primary and ships updates afterward to minimize write latency.

    Example

    An online ticketing platform runs a primary PostgreSQL database in us-east to process seat reservations while streaming write-ahead log (WAL) changes to two read replicas that handle high-volume event search queries. When a hardware fault takes the primary node offline, an automated failover controller promotes a standby replica to primary within seconds so ticket sales continue with minimal disruption.