Gossip Protocol
also: epidemic protocol
Nodes trade state with random peers; information spreads like rumor — no coordinator, log-time propagation.
In a gossip protocol, each node periodically exchanges what it knows with a few random peers, and updates spread epidemically through the cluster in O(log N) rounds. No central point, tolerant of failures, bandwidth spread evenly — ideal for membership, failure detection and metadata in Cassandra, DynamoDB and Consul. The trade: convergence is probabilistic and eventual, not instant.
Worked example: nodes periodically exchange state with a few random peers, so information spreads epidemically — each round roughly triples the informed set, reaching all N nodes in about log(N) rounds with no central coordinator. Gotcha: gossip is robust and scalable (no single point of failure, tolerates churn) but eventually consistent and chatty — it trades immediacy and bandwidth for resilience, which is why it powers membership/failure-detection (Cassandra, Consul) rather than data needing strong ordering.