---
title: "Cerberus (Consensus Protocol)"
path: "/contents/tech/core-protocols/cerberus-consensus-protocol"
bannerImage: "https://j9ytchrzkvqpcul7.public.blob.vercel-storage.com/334738ec-99ff-4992-a818-fe9ed5431b74.jpg"
version: "5.3.0"
author: "Hydrate"
createdAt: "2026-01-11T17:06:09.522Z"
updatedAt: "2026-03-16T18:25:44.340Z"
---

# Cerberus (Consensus Protocol)

<Infobox>
| **Type** | [BFT](https://en.wikipedia.org/wiki/Byzantine_fault) Consensus Protocol |
| **Whitepaper** | [2020 (peer-reviewed)](/contents/tech/research/cerberus-whitepaper) |
| **Key Innovation** | Braided cross-shard consensus |
| **TPS** | 500,000+ sustained ([Hyperscale test](/ideas/hyperscale-500k-test)) |
</Infobox>

## Overview

**Cerberus** is Radix's novel Byzantine Fault Tolerant consensus protocol. Its key innovation is **braided parallelism**: instead of running a single chain or fixed shards, Cerberus dynamically braids consensus across only the shards relevant to each transaction.

### How It Works

Each transaction identifies the [substates](/contents/tech/core-concepts/substate-model) it needs to read or write. Cerberus forms a temporary consensus group from the validators responsible for those substates' shards. The protocol uses a **3-phase commit** (pre-prepare, prepare, commit) to achieve atomic consensus across all involved shards.

Transactions touching different substates on different shards execute in parallel with zero coordination overhead. Only transactions with overlapping substates need to be ordered relative to each other.

### Linear Scalability

Adding more shards (and more validators) increases the network's total throughput proportionally. This is **linear scalability** — the theoretical throughput ceiling is unlimited. The [Hyperscale public test](/ideas/hyperscale-500k-test) demonstrated 500,000+ sustained TPS across 128 shards with 590+ nodes.

## Comparison with Other Consensus

**vs. [Ethereum](https://ethereum.org) PoS**: Ethereum processes all transactions sequentially through a single execution layer. Cerberus parallelizes across shards while maintaining atomic composability.

**vs. [Solana](https://solana.com) Tower [BFT](https://en.wikipedia.org/wiki/Byzantine_fault)**: Solana uses a single-threaded PoH + Tower BFT design requiring high-end hardware. Cerberus uses commodity hardware with throughput scaling via shard count.

**vs. [Cosmos](https://cosmos.network) [Tendermint](https://tendermint.com)**: Cosmos chains are sovereign with [IBC](https://ibcprotocol.dev) for cross-chain communication (asynchronous). Cerberus provides synchronous atomic composability across shards.

## External Links

  - [Cerberus Whitepaper](/contents/tech/research/cerberus-whitepaper)

  - [Academic Paper (arXiv)](https://arxiv.org/pdf/2008.04450)

  - [Hyperscale 500k TPS Test](/ideas/hyperscale-500k-test)