Eventual Consistency¶

When immediate consistency is impossible or prohibitively expensive across distributed nodes, systems can guarantee that all replicas will converge to the same state given sufficient time without new updates.

Tension Addressed¶

Consistency vs. Availability (accepts temporary inconsistency to maintain availability)

Recognizable As¶

Replication lag warnings; "read your writes" guarantees; version vectors; last-write-wins policies; anti-entropy protocols; gossip protocols

Canonical Sources¶

Vogels, "Eventually Consistent" (2008); Brewer, "CAP Twelve Years Later" (2012); Shapiro et al., "Conflict-free Replicated Data Types" (2011)

Transfers To¶

Database Design (replication strategies), Caching (invalidation timing), Distributed Filesystems (sync semantics), Collaboration Software (concurrent editing)

Eventual consistency acknowledges a physical reality: information propagates at finite speed. Rather than blocking all operations until consistency is achieved, the system makes progress and reconciles later. This requires designing for the inconsistency window: what happens when a user sees stale data?

Sub-concepts: - Read-your-writes Consistency: Guarantee that users see their own updates immediately - Causal Consistency: Related operations appear in causal order (if A caused B, everyone sees A before B) - Conflict Resolution Strategies: Last-write-wins, merge functions, manual resolution