Unitalk with Prof Sirer of Avalache: the New Simple Yet Powerful Consensus Mechanism

Originally posted on 24, Jan. 2019.

Consensus is the basis of all transactions. The more dispersed the consensus (the higher the participation), the lower the efficiency, but the higher the satisfaction, the more stable it is. On the contrary, the more concentrated the consensus (the lower the participation) The higher the efficiency, the more prone to dictatorship and corruption.

The great thing about blockchain technology is decentralization and de-trust (or lowering credit costs). How do you match transactions through code in a completely unfamiliar network environment? What is the meaning of the words, what magical power does the code play, and let the buyers and sellers generate trust?

The answer is the consensus mechanism.

The three most commonly used consensus algorithms in the blockchain are PoW, PoS and DPoS.

However, most of the current consensus mechanisms have scalability problems, which make the TPS and throughput of the blockchain difficult to support mainstream adoption.

In May 2018, Cornell University professor and computer scientist Emin Gun Sirer announced a new consensus agreement, Avalanche, at the Token Summit in New York. The agreement was created by an anonymous developer team called Team Rocket, and Professor Sirer and his team wrote the white paper.

Hi, Prof. Sirer. Nice having you with us. Can you introduce yourself and Avalanche first?

I’m a professor at Cornell University, whose career has focused on peer-to-peer technologies. Specifically, I was the first person to build a proof-of-work currency, in 2002/2003, some 6 years ahead of Satoshi. I then worked on improving the security of Bitcoin, predicted the DAO hack, and worked on new consensus protocols. I’m now also the cofounder of bloXroute, as well as the founder and CEO of Ava Labs.

Now, as for Avalanche, Avalanche is a big breakthrough in distributed systems. For the first time, we have a consensus protocol that can truly scale, and support features that no other protocol can support. It can scale not just in terms of transactions per second, but also in the number of people who can participate first hand in the system. Anyone can join Ava, help secure the system, and get rewarded for doing so.

To put this in perspective, all past work over the last 40 years in distributed systems can be categorized into just two consensus families: classical protocols, and Nakamoto consensus. Avalanche is the third big invention in this space in over 40 years, a 1-in-15-years kind of event.

What inspired you to create the new consensus protocol? What pain point are you trying to solve?

We, the crypto community, sold a fantastic dream to the masses. A compelling dream that has been unreachable with the technology that we possessed until recently. Today’s cryptocurrencies cannot reach the masses. They don’t scale — even Bitcoin developers admit this and are looking for solutions at a different layer. They consume immense amounts of energy, which causes value to leak out of the Store of Value. Mining is inherently centralized, and cuts out ordinary people from the process of maintaining the ledger.

Ava addresses these problems. It offers unprecedented decentralization and performance, along with new features not realizable with existing chains. Whereas many blockchains admit only about a 100 miners or block producers to participate in the system, Avalanche consensus allows for thousands while maintaining very high throughput.

How long does it take to achieve finality across the globe using Avalanche? What is the throughput?

It takes about 1–2 seconds for Ava to achieve finality. The throughput that the system can achieve in our benchmarks is around 10,000 TPS on chain.

These are numbers from geographically distributed benchmarks, not just synthetic results from the lab. These benchmarks were collected on 2000 nodes deployed across the globe.

The core idea of Avalanche is metastability. Can you explain to us in real-world analogy what metastability is? What it has to do with Avalanche?

Metastability is a concept from physics, a phenomenon often seen in dynamical systems of interacting particles. In such systems, particles may be in a mixed set of states, but if coaxed and cajoled properly, they can quickly settle into a stable form where all particles share the same state. Some of you may have seen, perhaps in a high-school chemistry or physics class, crystals settle out of the solution. Metastability is about finding global states of lower energy, where the particles undergo a phase shift and form an unchangeable, stable structure. It is the process of dynamic chaos, settling down into a static, stable state, exactly what we want to have happen in a distributed system.

Avalanche consensus is heavily inspired by this concept. In a consensus protocol, you want everyone to agree on the ledger contents. A metastable algorithm guarantees that the system will not vacillate or flip flop — it’ll settle quickly and remain immutable. Avalanche is the first consensus protocol to be designed around the idea of metastability. The use of this new mechanisms completely changes the game on consensus protocols. What used to be costly and fragile can now be much more efficient and robust.

Does Avalanche have a token? What are its properties？

Avalanche will have a token. The token will serve as a means of transaction. The sybil control mechanism will be proof-of-stake and so participants will be able to stake tokens in order to validate transactions.

The token also enables the holders to vote for improvements in the protocol and ecosystem.

Token holders can have a first-hand say on key system properties, such as the rate at which tokens are minted. They can use their tokens to vote on such governance issues.

Will AVA be able to interact with other chains?

Indeed, we are the first coin to commit to supporting multiple virtual machines. We currently support both the Bitcoin scripting language as well as our own native VM, and we have plans to support EVM and WASM.

In addition, other coins will be able to use Avalanche as a pre-consensus or post-consensus mechanism. For example, proof-of-work currencies can use avalanche to agree on the transactions that go into each block. Or they can use Avalanche to automatically establish checkpoints to protect themselves against 51% attacks. This would be in addition to the security provided by PoW.

I read that if somebody misbehaves, he won’t lose his tokens. So what is the penalty of malicious behaviors?

Indeed there will not be any “slashing” in the protocol. The avalanche consensus mechanism naturally prunes out malicious behaviors, and the stake required to attack the network provides the necessary protection.

In contrast, classical consensus protocols have to necessarily punish malicious behaviors, because they can be attacked essentially at 0 cost. As a result, running a staking node with protocols other than Avalanche is dangerous. If there is a hardware or software failure, you can lose your stake with those protocols.

We believe in quiescent, calming protocols. One’s coins should not be at risk. Even hardware failures should not cause you to lose your coins.

That’s why Avalanche does not use the “slashing” technique. A staker will always be able to get back their stake, and if they participated in the protocol, they will also get some number of freshly minted tokens to compensate them for securing the network.

Then how does Avalanche prevent double-spending attacks or other malicious behaviors?

The Avalanche protocol is naturally good at preventing double spends. How Avalanche works is that when there is a double spend, the network runs a repeated sampling protocol to determine which of those transactions is valid and adopted by the rest of the network. In essence, every participant repeatedly selects a small number of other participants and asks them which of the two transactions they believe is the one to adopt. They then switch their own behavior to pile behind the dominant transactions, building and evolving a DAG around the accepted transactions. The metastable nature of this repeated sampling and building process guarantees that the network will adopt only one of those transactions.

I learned that you identified multiple flaws in The DAO and urged a moratorium prior to the hack. It was not your job. Why did you do that?

Ah, that’s a great story. I have been working to improve the space overall. As a scientist, I see it as every bit of my job to not just come up with new protocols, but to ensure that the industry is doing the right things. If people lose their investments, and there is heartache and bloodshed, it’d have terrible ramifications for the industry.

When I saw that the DAO had collected in excess of $200m in funds, it was clear to me it had become a target of attack, and that a failure could have terrible consequences for not only Ethereum but also the entire space of cryptocurrencies.

Vlad Zamfir was visiting my research group at the time, and he pointed out the DAO was in a prime position to be exploited.

So we started looking carefully into the security of the contract. We ended up fighting against the clock: the contract was programmed to go live, so we had to finish our work and warn the public. We went public with 6 flaws, put together in a google doc.

The Google document went viral, and 100s of people were viewing it, as we were extending it in real time. I remember typing up 3 additional flaws, with countless eyeballs viewing the document as I typed.

The document was titled “A Call for a Moratorium on The DAO.” The community was debating what to do about the DAO, but of course, it was difficult to stop. Then the hacker came in, and the rest, of course, is history.

What’s your opinion on the postpone of Constantinople hard-fork? Do you have anything to say to the Ethereum community?

I haven’t followed the recent Constantinople issue too closely, but I think it was the best thing to be prudent. From what we understand, the bug relates to reentrancy. I explain what this is specifically in my blog in the context of the DAO hack.

You can check it out here.

I have great respect for the Ethereum community because of their commitment to science and their positive attitude. Keep up the good work!

Vitalik Buterin: It’s immutability stems from the metastable nature of the core algorithm. But aren’t traditional BFT algos also metastable in the same way? All the ones I know about definitely have the property that if a strong majority is voting for one value they will stick to continuing to vote for the same value.

Hi Vitalik! The answer to this is no. Traditional BFT mechanisms can remain in the bivalent state forever, and can remain stable in the bivalent state forever.

Punishment is an essential feature of any cryptosystem. Why did you omit that part in Avalanche?

Punishment is an essential feature of any classical consensus protocol! That’s because in a classical protocol, if misbehavior does not have any penalty associated with it, people can get away with it and revert transactions. In contrast, Avalanche does not need the punishment feature to ensure immutability. Its immutability stems from the metastable nature of the core algorithm. So, because we do not need to punish, we do not, and save everyone the anxiety associated with slashing conditions.

What do you think needs to happen to accelerate the adoption of cryptos by non-technical people?

First of all, we need to get new protocols that can actually live up to the dream that has been sold to the masses. None of the current protocols can support even a small country at the moment. If Venezuella adopts BTC, every adult gets to transact once every 36 days. Ethereum gets bogged down just by cryptokitties. We cannot reach the mainstream with this infrastructure. So the most important thing right now is to find new infrastructure that can scale. That’s the first and most important impediment.

I also worry about the user experience. Hardware wallet usability leaves much to be desired. Integration with the regular financial system, and with pegged stable coins, are all critical. And of course, new applications targeting underserved industries.

We know Nakamoto protocols such as POW can achieve probabilistic consistency and classical BFT protocol such as PBFT can achieve strong consistency. It’s always a conflict to realize BFT consensus on public chain due to it’s O(n²) communication complexity. Could you explain what kind of consistency can Avalanche achieve? How and by what to realize this kind of data consistency?

Indeed, classical protocols provide what we call “finality with probability 1.” But to achieve this guarantee, they make lots of very strong assumptions, such as assuming that everyone knows everyone else who is part of the system. This in itself is a very difficult condition to ensure in large networks, and renders these protocols quite fragile. They also typically require N² communication.

The genius of Satoshi was to look at this and say “hey, I don’t need to give you probability 1. I’ll give you probability 1-epsilon. The two are pretty much the same thing if epsilon is small.”

Avalanche is in the latter category: it provides a probabilistic guarantee of safety (consistency). But the probability is essentially 1. The chance of a safety violation is much lower than the chance that some alpha ray from the sub will hit your CPU and will cause it to send your funds to a wrong address.

Why there are so many protocol projects in IC3, like Link, Thunder, Avalanche, Bloxroute. Are these projects related?

Because we are very productive! :-) And it’s not me. I’m proud of the super bright graduate students who are part of IC3. They do all the heavy lifting. And of course, it helps to have the smartest minds so closely coupled together. They interact, and spur each other on.

The projects you mentioned are not related, but they all interlock and share common insights.

What do you think about current DEX technology? What should DEX focus on in order to have mass community adoption?

I love DEXs!

If I may go on a tangent: the Bitcoin community repeatedly goes in front of the SEC and tries to get an ETF approved. Every two months, there is a new Bitcoin ETF proposal before the SEC, and *every single time* the SEC has the same response: “the existing exchanges for crypto are highly manipulated, and therefore an ETF cannot be approved at this time.” For some reason, people then try the same thing over and over again, without learning or changing anything. The fiat folks keep snubbing us.

Except the funny thing is that, when you look at fiat exchanges, they are total messes themselves that require layers upon layers of audits to make work right. Because of the nature of our cryptoassets, we are **in a position to build much more trustworthy exchanges than the fiat folks themselves.** We can lead Wall Street, show them how trustworthy DEX’s can be constructed. We can turn to the SEC and say “no, our exchanges are secure, and you all need to up your game and adopt our technology.”

I can’t wait for that day, and I look forward to building and deploying such trustworthy systems on top of Ava.

ou are always a supporter of big blockchains. But it is obvious that the big blockchains have been defeated in the competition. Their personal interests influence the consensus of Bitcoin, such as BCH. Do you believe that the price of BCH is manipulated by someone to remain at a relatively high price?

I take pride in being principled and scientific. I have criticized big blockers as well as small blockers. If I seem to be on the big blocker side in the blocksize debate, it’s because I saw the small blockers use arguments and social media manipulation techniques that I found to be at odds with scientific reasoning. To be fair, big blockers have also engaged in some bad practices, and I’ve called them out on it as well. I would very much like us to evolve towards sound, trustworthy systems, proven by science.

As for price manipulation, it’s not something on which I have any unique views. There are tons of coins in the top-10 that have market caps *far* in excess of their technical merits. I believe both BTC and BCH are exploring two reasonable paths, one if exploring scaling offchain while the other is trying to push the original design to the maximum allowable by the networks and hardware.

You mentioned that if a dishonest node attempts to implement a double-spend, the Avalanche protocol will not be able to reach a consensus on either of the operations. Do you view this as an advantage or disadvantage?

Let’s be clear, because this is a bit nuanced, and there are two separate issues: if a dishonest node attempts a double-spend, Avalanche *will* decide.

But if a dishonest node is coupled with a very large-stake attacker, he may be able to keep the network from picking one of the two transactions. This has no negative impact on Ava, as the system builds a graph (a DAG), and it will simply leave those two tx’s in limbo, and grow the graph around them.

Know more about Unitimes:

Website: www.untimes.pro / www.unitimes.io

Twitter: UnitimesHQ

[The copyright of this article belongs to Unitimes. Please contact us at editor@unitimes.io if you want to repost the article. Opinions expressed by contributors belong to themselves.]