On 16th of May 2018, an unknown group published a paper claiming a very exciting scientific development called Metastable Consensus protocol. Before this in the past 50 years there have emerged only 2 consensus family called classical consensus protocol and Nakamoto consensus protocol.
Emin Gur and his team have been working on this for last 2 years, but sometime around february this year they were contacted by an anonymous group who gave Emin and his team the backbone of this paper they published. They essentially scooped Emin and his team on what they have been trying to do in the past 2 years.
He states that the protocol is simple and powerful. It does for distributed system in consensus what gossip protocol did for data dissemination. In the old days when you have something to say and when you wanted to send this across to a bunch of people, you would use what we call a dissemination protocol and again in that time people who built this protocol had rigid structures in mind, more like a distribution tree with different tiers where message flow between different tiers. But this system is inherently fragile where everybody have to know about the tree and everybody have to have to understand where they were on the tree; what their role role was and who their decendents were.
At some point at early 90’s people came up with gossip protocol where essentially all you do is you pick a couple of people at random and you tell them what you know and this is an incredibly powerful mechanism for disseminating information. Because this random networks are incredibly robust, you don’t have to know everybody in the system, you just have to know some people and you have to communicate only with them and yet in logarithmic time you get whatever you have to say across to the entire network.
So this new consensus protocols known all named after various forms of snow namely: Avalanche, Snow Flakes, Snow Ball. They rely on the randomness and random interaction and yet they ensure that at the end of their interaction everybody has decided on the same thing, and that is a very powerful primitive. At the core of this algorithm there are a series of ideas that they invented that are very powerful, it’s a new way of doing consensus.
According to Emin he suggests anyone who is working on anything related to consensus have to stop what they are doing and take a look at this new consensus protocol, as the protocol really promises to revolutionize how one can do consensus for this domain. He is not sure if this will replace proof of work system as he believes that there will be people who really like proof of work system in which case proof of work system can use avalanche protocol for miners to talk to each other and agree on the contents of the next block, whereby making use of avalanche low latency on proof of work ecosystem.
The way this works is:
If we all were to agree on something for example picking between colors red and blue. In case of Nakamoto consensus we would have miners actually mine before coming to consensus and it would be very costly and as well coin minting would be expensive as a lot of coin needs to be minted to reward the expensive proof of work operation. Whereas if we were to implement a classical consensus protocol where everybody asks everybody else to reach 2/3 plus 1 majority and to reach that majority that node would have to actually know everybody in the system and that these systems are very fragile and that’s why one of the reasons why they have not been deployed at scale. But in the case of this new consensus protocol family what you can do is simply pick some numbers of people for example 5 people, and ask what color they prefer? The node will change its color to what majority in that 5 decided, and this would iterate over and over until a constant number of iteration when all participant nodes end up with the same color, which is quite powerful and quite unexpected.
The reason why it works is incredibly simple as well. Imagine a worst case scenario of a network divided 50 50. We all pick 5 people, there will be random variation, and at the end of the first round we would have over sampled ether blue or reds. So at the end of the 1st round there would be slightly more red or slightly more blue nodes, and second round we will amplify and the third and suddenly we will reach a point where we all will fall to one extreme where we are either all red or all blue. This is what is called Metastable Protocol which means its designed to to not stay balance which is exactly what we want out of consensus protocol and which always fall one side or the other.
Because this protocol is light weight with super low latency of around 1-2 secs and many thousand transaction per second, Emin and the team is planning to support other existing virtual machines such as Bitcoin and Ethereum(EVM).