Developer Resources

Achieving Consensus

Find out how consensus is achieved on the Internet Computer and how long the asynchronous finalization of the consensus protocol takes

IC Academy » Developer Resources » Achieving Consensus

Consensus Protocol

Learn more about the Internet Computer’s Consensus Protocol

Asynchronous Finalization of the IC Blockchain

Reliability and security are the two key features of the Internet Computer blockchain. The network is powered by node machines hosted in data centers all across the world. The Internet Computer Protocol (ICP) is used by the nodes for communication. In doing so, a virtual Internet Computer is formed which allows developers to develop and deploy canister smart contracts in a reliable and secure manner.

In the context of the Internet computer, the term “secure” means that a canister’s state only changes based on the rules defined by the developer. This makes it impossible to temper with the canister. Secondly, the term “reliable” indicates that the canister that’s not stop running unless stopped by the developer.

The aforementioned properties of reliability and security are accomplished by utilizing a novel consensus protocol.

Overview of the Internet Computer’s consensus protocol

The network’s node machines (or replicas) must reach consensus on the state of the blockchain. This includes reaching consensus on which data inputs to process and on the order for doing so. Instead of utilizing just one node, many node machines execute software. As such, the true state of the blockchain is defined by the majority of the nodes. As long as the software is correctly executed by the majority of replicas, issues with individual nodes (such as connection problems, maliciousness, or reporting a tempered state) do not make a difference on the consensus mechanism.

Additionally, the Internet Computer is designed to scale indefinitely, allowing an unlimited number of decentralized applications to run on the platform. The network’s secure scalability is accomplished by utilizing the Internet Computer’s peer-to-peer layer.

To achieve scalability, the Network Nervous System splits nodes into subnets that run a given amount of canisters. Many different replicas power a subnet and execute the subnet’s canisters.

The Internet Computer can seamlessly grow its capacity by adding new subnets and by splitting existing ones. Furthermore, Chain Key Cryptography enables secure communication between canisters hosted on different subnets.

The Importance of Consensus

A consensus protocol is required to utilize replication for increasing the network’s security. This allows the correct state of a subnet to remain unaffected should replicas become malicious or unavailable – as long as the majority of node machines continue to function correctly.

These are the requirements for a subnet:

Subnets need to be able to process different messages (sent from users to canisters and canisters to other canisters)
Messages need to be processed in the same order to ensure that replicas end up in the same state

The problem lies therein that each replica in a subnet may see all the messages in a different order.

To overcome this problem, a consensus protocol is used to ensure the subnet’s nodes reach agreement on the order of messages to be processed.

Consensus properties

The consensus protocol places messages in blocks. Agreement is reached using a blockchain.

Block x

Block x+1

Block x+2

For the purpose of this example, we assume there are n replicas running a subnet together

Safety

If two (honest) replicas think that the i-th block is agreed upon, they must have the same block

Liveness

At some point, every (honest) replica with think that the i-th block is agreed upon, for any i

Validity

All agreed-upon blocks are valid

We want this even if up to f of the n nodes misbehave, with n = 3f+1. In the example above, we use n = 4, f = 1.

To reach consensus, a blockchain is used. Messages that need to be processed by a subnet are grouped and then placed in blocks. Each block is pointing to a previous block, forming a connected chain of blocks. If all replicas agree on the state of the blockchain, security is achieved as this provides a message ordering.

The Four Main Parts of the Consensus Protocol

The design of the consensus protocol is tailored to the specific needs of the Internet Computer network. Let’s have a look at its main parts:

Block making

Creation of candidate blocks used to build blockchains

Notarization

Identifying valid blocks that can be used to build valid blockchains

Random beacon

Randomness used to further enhance the protocol

Finalization

Information about having reached agreement

Block making

A replica within the subnet serves as block maker. It proposes a new block which extends the existing chain. The replicon gathers incoming messages and groups them into a block. The block is it then sent to the subnet’s other replicas with a proposal to extend the blockchain with it.

Many replicas is within the subnet can serve as block makers to ensure that the consensus protocol continues to work even if nodes are misbehaving. For this reason, it is not possible to elect one single block maker.

Notarization

Block proposals may be invalid because there is not a single block maker. Notarization ensures that there is at least one valid block to extend the blockchain.

Complications may arise if a replica creates a notarization share on a block but identifies another valid candidate block at the same block height. If only one block is signed, it could cause the chain to get stuck.

To ensure the liveness property is adhered to, the notary can sign both blocks so that at least one of the blocks becomes notarized.

Random beacon

Agreement has not yet been reached as multiple blocks may have been notarized at every height. The number of notarized blocks is therefore reduced by adding the random beacon to the protocol. It is an unpredictable value that is added at every height as a random-looking artifact.

With the help of the random beacon’s common randomness, the block makers are ranked with every round.

Finalization

To observe agreement a separate asynchronous finalization process is used. It lets the replicas know that they can accept the chain. With the Internet Computer’s finalization approach, blocks can be finalized in less than two seconds.

Full finalization implies that there cannot exist another notarized block – under the condition that less than one third of the subnet’s nodes are corrupt.

Summary

The consensus protocol of the Internet Computer is comprised of four key components. The first component is the block maker that extends the blockchain by creating candidate blocks. The second component is the notarization process for ensuring the correct identification of valid blocks. Thirdly, the random beacon is added so that the block makers can be ranked, leading to a reduction of notarized blocks in every round. The fourth and last component is the asynchronous finalizaton mechanism for informing nodes when a blockchain is agreed upon.

Interface Specification

Learn more about the Interface Specification that details how users and services communicate

IC Internet Specification

Introducing the Quill Toolkit

Find out more about Quill, the open source ledger and governance toolkit for the Internet Computer

Quill