> For the complete documentation index, see [llms.txt](https://opengpu-network.gitbook.io/opengpu-network/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://opengpu-network.gitbook.io/opengpu-network/blockchain-reference/consensus.md).

# Consensus

OpenGPU implements a consensus system designed for rapid transaction finality, high throughput, and robust security in decentralized applications. The architecture combines Asynchronous Byzantine Fault Tolerance (ABFT) with Directed Acyclic Graph (DAG) structures, drawing inspiration from the Lachesis protocol.

## Lachesis Protocol Foundation

The system builds upon Lachesis, an ABFT consensus engine originally created for high-performance, fault-tolerant DAG-based blockchains. Key inherited advantages include:

* Validators operate independently without requiring real-time coordination
* The consensus model eliminates reliance on leaders or synchronized global time
* Network maintains stability with up to one-third of nodes behaving maliciously

This foundation enables validators to achieve agreement through asynchronous communication rather than sequential voting rounds.

## Asynchronous Byzantine Fault Tolerance (ABFT)

ABFT operates by allowing nodes to work independently without real-time synchronization. The process involves:

1. Validators independently verify transactions
2. Event data exchanges happen asynchronously across the network
3. Consensus crystallizes once sufficient validators observe and accept identical data

**Benefits:**

* **Scalability**: Parallel validator operations increase throughput
* **Resilience**: Sluggish or faulty nodes don't impede network progress
* **Speed**: Transactions achieve finality within 1 second

## Directed Acyclic Graph (DAG) Architecture

Rather than employing linear blockchain structure, OpenGPU uses DAG representation to model event relationships:

* **Vertices** represent event blocks containing transactions
* **Edges** show references and dependencies among blocks
* **Concurrency** allows validators to generate and propagate blocks independently

This structure enables substantially higher throughput and faster confirmations versus traditional sequential blockchains.

## Event Block Lifecycle

The progression follows five stages:

1. Users initiate transactions
2. Validators create event blocks and append to local DAG
3. Asynchronous peer-to-peer block sharing occurs
4. Blocks achieving majority observation become "root events"
5. Root events finalize into the canonical "main chain"

The resulting main chain represents the definitive sequence of finalized transactions accessible through block explorers.

## Security & Finality Guarantees

* Finality typically occurs within 1 second
* Architecture tolerates 1/3 malicious nodes maximum
* Proof-of-Stake mechanisms prevent Sybil attacks while ensuring economic incentive alignment

## Summary

The synthesis of Lachesis-inspired ABFT with DAG efficiency produces a consensus framework delivering:

* Elevated throughput paired with rapid transaction settlement
* Network resilience against latency and adversarial participants
* Scalable validation suitable for contemporary blockchain applications