Consensus Algorithm
Subject: Computer and Technology noun /kənˈsɛn.səs ˈæl.ɡəˌrɪð.əm/ 1 definition
- distributed-systems
- fault-tolerance
- state-machine-replication
- distributed-computing
-
Definition (noun)
Entry #71 · UpdatedA Consensus Algorithm is a fault-tolerant protocol used in distributed computing to help multiple independent nodes agree on a single data value, execution order, or shared system state across an unreliable network. Because participants can experience variable message delays, network partitions, hardware crashes, or adversarial behavior, the algorithm defines strict proposal, voting, and commitment rules so that all correct nodes maintain a consistent view of truth. Rather than relying on a single permanent central server, it coordinates decentralized agreement while balancing safety—guaranteeing that honest nodes never commit conflicting results—against liveness—ensuring the cluster continues making progress whenever enough participants remain reachable.
Common Types
Crash Fault-Tolerant (CFT) Consensus: Protocols such as Paxos and Raft that assume nodes may fail or lose messages without acting maliciously, reaching deterministic agreement whenever a strict majority quorum remains reachable.
Byzantine Fault-Tolerant (BFT) Consensus: Algorithms such as Practical Byzantine Fault Tolerance (PBFT) that preserve agreement even when fewer than one-third of nodes fail arbitrarily or behave maliciously, relying on multi-round signed message exchanges among known validators.
Proof-of-Work (PoW) Consensus: A probabilistic, permissionless mechanism popularized by Nakamoto consensus in which participants expend computational power to solve cryptographic puzzles for block-proposal rights, converging on the chain with the greatest cumulative work.
Proof-of-Stake (PoS) Consensus: Protocols that assign block-proposal and voting weight in proportion to locked cryptocurrency collateral, deterring equivocation and conflicting votes through economic slashing penalties.
Example
A five-node etcd cluster backing a Kubernetes control plane uses the Raft consensus algorithm to replicate cluster configuration across all members through an elected leader. When one follower crashes or becomes isolated by a network partition, the remaining majority quorum keeps committing updates safely, and Raft automatically synchronizes the lagging node's log once it reconnects.