OPRR Flash Depth Activities

Search Extension

Avalanche DEX developer explains Avalanche consensus mechanism

2021-12-14 18:17

Read this article in 12 Minutes

AI summary

View the summary

Connor Daly is the founder of Avalanche DEX Pangolin

Eternal, BlockBeats

This article is from Connor Daly, founder of Avalanche DEX Pangolin, on his personal social media platform, translated by BlockBeats as follows:

Avalanche has revolutionized every layer of the blockchain network, but it has one innovation above all others. When I found out, I quit my job and dedicated myself to Avalanche.

Here's the definitive explanation for Avalanche.

First, what is consensus?

It is a mechanism to ensure that multiple nodes in a distributed system make the same decision, and for a blockchain, the most important decision is to decide which block should become the new block on the chain.

Before Avalanche's consensus, there were two major sets of consensus algorithms: Classical and Nakamoto.

Classical's consensus was achieved through multiple rounds of voting. Each node in the network communicates with other nodes to broadcast its decision. Since nodes need to communicate with other nodes in the entire network, members of network nodes need an accurate list of protocol members.

The Classical consensus algorithm works well with a small number of nodes, but its time complexity O(n^2) is particularly high, resulting in high operating costs for the node network (often proportional to the quadratic of the number of nodes N). At the same time, this algorithm cannot be extended to large decentralized networks that support user mobility.

Some blockchains using Classical consensus algorithms have tried to mask performance deficiencies. These blockchains often deceive users by allowing only a small number of validators to participate in the consensus mechanism and block validation process. In such blockchain networks, there are ostensibly 1,000 verifiers, but in reality only 100 participate in consensus decision-making.

The second kind of consensus algorithm faction is Nakamoto consensus. That's what bitcoin and many other cryptocurrencies do. It uses the "longest chain principle", assuming that the chain that consumes the most work has an authoritative historical record. Miners do not participate in voting verification, but hash and distribute their work.

The advantage of the Nakamoto consensus algorithm is that it can scale to a large number of users without sacrificing performance. The problem is, performance isn't great anyway.

Any number of miner nodes can participate in or exit the block blasting process at will, which leads to slow block generation and block reorganization at any time, leading to changes in the storage contents on the chain.

You may have heard that a transaction on the Bitcoin network was finalized after six new blocks were created. But this is only a 99 percent high probability event, based only on confidence estimates.

The more times a new block and its chain are identified by network nodes, the less likely it is that a longer chain fork will replace it. But that doesn't guarantee that longer chain forks won't occur.

Avalanche's consensus mechanism. This is a brand new consensus algorithm, which is realized by random subsampling.

The consensus protocol was invented and standardized by researchers at Cornell University, including Avalanche co-founders Emin Gun Sirer, Ted Yin, and Kevin Sekniqi.

Let's use an example to illustrate the Avalanche consensus mechanism.

Suppose there are 1,000 meeting participants who need to decide together what to eat for lunch: pizza or sandwiches. Everyone has an initial preference, but they need to agree on the final plan.

As a participant, you take a small random sample and ask 10 participants what they want to eat. Seven people say pizza. It looks like most people want pizza, so now if someone asks you, you'll say pizza.

Next, you perform another round of random sampling. You ask another group of 10 randomly selected people, and if most of them answer sandwich, you change your preference; And if most people say pizza, you'll say pizza again. So you get the same result from both samples.

You can continue iterating through this process until you get the same result 100 times in a row. At this point, you think the final decision is made.

Since nodes only communicate with a fixed number of nodes in each round of sampling, the time complexity of message delivery in a single round of sampling is O(1) (often at a constant level). Unlike the Classical consensus algorithm, the complexity of message passing does not increase as the number of nodes N increases.

But how many rounds of random sampling are needed to get the final result? It depends on the size of the network. However, its complexity is a controllable O(log n) (proportional to the logarithm function, order of magnitude lower than the first order function).

The avalanche consensus is based on probability statistics. A node may repeatedly sample only a small number of nodes and draw incorrect conclusions. But the protocol is tunable. We can make the probability of this happening extremely small, such as less than 0.0000000001%.

Because of this high degree of predictability, Avalanche is final. Once intra-network transactions are decided, they are locked on the chain forever. There will be no block reorganization at Avalanche.

This is a visual representation of Avalanche consensus.

https://twitter.com/das_connor/status/1470147125870841861

If you look carefully, you will see that after certain nodes are selected, the final result of multiple rounds of sampling is impossible to reverse. It's like a moving snowflake turning into an avalanche. The Avalanche protocol gets its name from Avalanche.

Now that you know how it works, why is it important? What can it do?

Attribute 1: The protocol has high operating efficiency, and the limiting factors are message delay and network bandwidth.

This means you can plug into the Avalanche network on low-spec hardware, without the need for specialized mining equipment, just servers in the middle tier. Some users even run Avalanche verification nodes on Raspberry Pi microcomputers.

Property 2: The protocol is static, running only when there is work to be done, without the "waste" of a proof-of-work (POW) mechanism.

Attribute 3: Avalanche's consensus algorithm is environmentally friendly. Because the hardware requirements are very low, the impact on the environment is minimal. Avalanche also recently partnered with entrepreneur Vinay to make it carbon neutral.

Attribute 4: Avalanche consensus algorithm is fast to execute, Avalanche can process 4500 transactions per second and finalize in 1-2 seconds in simulation and practice.

Attribute 5: Avalanche can use any form of Sybil protection mechanism. You may have heard that Avalanche is a proof-of-stake (POS) blockchain. But POS is not a consensus algorithm, it's just a protection mechanism against Sybil attacks.

The Sybil protection mechanism prevents a single user from manipulating multiple nodes and masquerading as multiple users. To run multiple Validators, you need to have a large stake, which we consider a scarce resource.

Although the Avalanche protocol requires Sybil protection, you can use any of the other methods. For example, you can rebuild Avalanche using proof of work (POW).

Avalanche consensus is configurable (the system can be configured to meet operational needs). It is essentially a new consensus protocol with no fixed code implementation.

For example, you can choose to set network parameters to prioritize security or activity.

With Avalanche Blockchain, we've tweaked the network to make it highly secure. Unlike the "51 percent attack" on Bitcoin, hacking Avalanche requires 80 percent of the nodes to collude.

Bitcoin's "51% attack" looks like a centralised organisation falsifying history and suppressing minorities. At Avalanche, 80% of the nodes conspired to cause security failures, causing nodes in different parts of the network to make different decisions. Avalanch balances security with activity. If one-third of the network nodes are offline, the network process will slow down or stop until it recovers to a threshold level.