Multi-Leader Replication

Also known as master-master or active/active. More than one node accepts writes.

Use Cases

1. Multi-Datacenter: A leader in each datacenter. Tolerates DC outages and gives better local write performance.
2. Offline Clients: Devices (phones, laptops) act as local leaders.
3. Collaborative Editing: Google Docs, Etherpad.

Conflict Avoidance

The simplest way to deal with conflicts is to avoid them. If the application ensures that all writes for a particular record go to the same leader, conflicts cannot occur.

• Example: Route all requests from a user to the same datacenter (the "home" DC) based on their location.

Convergence

If conflicts do happen, the database must resolve them in a way that all replicas arrive at the same final state (Convergence).

• Last Write Wins (LWW): Uses timestamps. Risky because clocks are unreliable.
• Version Vectors: Used in leaderless systems to track the history of changes across multiple nodes.