Byzantine Fault Tolerance

Introduction

Byzantine Fault Tolerance (BFT) is the ability of a distributed system to reach consensus correctly even when up to one-third of participating nodes are faulty or malicious. The term originates from the Byzantine Generals Problem, a thought experiment by Lamport, Shostak, and Pease (1982) about coordinating action among parties who may include traitors.

In blockchain systems, BFT ensures that the network can agree on the state of the ledger even if some validators are offline, sending conflicting messages, or actively attempting to disrupt consensus. BFT protocols guarantee safety (no incorrect state) as long as fewer than 1/3 of participants are Byzantine.

Classical BFT and Its Limitations

Classical BFT protocols like PBFT (Practical Byzantine Fault Tolerance) run a single consensus pipeline where 2/3+ of validators must agree. While provably secure, this creates a throughput bottleneck — all transactions flow through one pipeline, and the communication complexity grows quadratically with the number of validators (O(n²)).

This is the core tension in blockchain design: proven BFT security comes at the cost of limited throughput. Many Layer-1 blockchains "solve" this by reducing the validator count or using probabilistic finality, trading security for performance.

Cerberus: Parallelised BFT

Radix's Cerberus protocol takes a different approach: it takes a proven single-pipe BFT consensus process and parallelises it across a practically unlimited number of shards. Each shard runs its own local BFT consensus to order transactions on its substates.

When a transaction touches state on multiple shards, Cerberus "braids" consensus across those shards using an atomic commitment protocol. This ensures that either all shards commit the transaction or none do — preserving atomic composability across the entire network.

This design provides linear scalability: throughput increases proportionally as more nodes and shards are added. The Cerberus whitepaper demonstrated that the protocol can handle millions of transactions per second in simulation, outperforming other sharded consensus approaches while maintaining the full security guarantees of classical BFT.

External Links

• Byzantine Fault — Wikipedia
• What is Cerberus? — Radix Knowledge Base
• Cerberus Whitepaper — arXiv