Not every day that somebody in the field of distributed systems, comes out and says “I have a new breakthrough to tell you about”. However, recently, Prof. Emin Gün Sirer, from Cornell University, said that he does. In this article, we are going to describe to you a new family of consensus protocols that just emerged several months ago. “Team Rocket — A pseudonymous team — with the collaboration of Prof. Sirer dropped this paper titled “Snowflake to Avalanche: A Novel Metastable Consensus Protocol Family for Cryptocurrencies” on IPFS. The paper builds on comprehensive low-level details and proofs that are not easily comprehended by non-academics. In this article, we will simplify and visualize the protocol for you, leaving you with an understanding of this new approach.
Consensus is one of the most important goals to be achieved when many distributed computers share the same task and resources.
A typical example is maintaining account balances in a banking system over multiple servers. Banks don’t rely on one single database to carry their data, but many geographically distributed machines across the globe to handle these transactions. They require a method to maintain a stable view across all these machines, thereby, reflecting a consistent representation of account balances.
Another example is maintaining a consistent view of Amazon’s online shopping services. In order to scale these services globally, they have to be (geographically) distributed across large datacenters. Amazon requires having a regular view of its stocked products or else, unexpected and undesired behaviour might happen, e.g., the last item can be sold twice to two different people.
This has been an important problem in Computer Science for a long time. To this end, academics/engineers have been working hard over the course of the last 40 years to find efficient solutions. In the field of distributed systems there are only two existing families of consensus protocols which we describe next.
The first class called classical consensus protocols were developed by two great computer scientists, namely, Leslie Lamport and Barbara Liskov. Both of them are Turing award winners — which is the equivalent of a Nobel prize for Computer scientists. The advantages of this protocol are various including quick finality and quick guarantees about the committed transactions. However, this comes with a price:
Basically, their security comes down to the judicious overlap between a quorum of nodes that confirm seeing the same thing and go ahead commit to it. This approach is reasonable when building a permissioned blockchain, however, it is not a good foundation when you are in a dynamic environment of untrusted nodes. Therefore, permissionless chains are based on a different protocol for consensus.
In 2009, another new family appeared on the scene. Satoshi Nakamoto came up with his new protocol family and presented his solution that was uniquely robust:
However these benefits came with a price, in particular:
Avalanche comes in and combines the best of both world’s, in particular,
The core idea of Avalanche is metastability. The worst thing for a consensus protocol is to be unable to decide between two choices, i.e., you don’t want it to say either this happened or that happened you want it to fall to one of the two sides. Avalanche’s metastable protocol is designed to tip the choice towards one of the sides.
To explain the core intuition of the process, lets go through a simplified example, as visualized in the figure above. Consider a network of trustless nodes that want to vote for either blue or red.
The protocol resembles a recurring subsampled voting process. What happens here is that even if we start out in the worst possible scenario of a 50–50 split between red and blue, after one round, with high probability, the scenario will no longer hold. Furthermore, chances of that are astronomically small after two rounds, even smaller after three, and so on, i.e., it decays exponentially. The protocol is designed to tip and not stay in the middle. As it tips more and more, the network’s perception shifts to one color or the other. The speed with which we move towards one direction (nodes vote one color more than the other) is going to increase and increase and at some point we reach to the point of no-return where the entire network has agreed on a color.
AVA — short for Avalanche, resembles the new token represented by this protocol.
People often talk about the consensus protocol used in Bitcoin as if it was a pure decentralized approach that determines the opinions of all network participants equally. Unfortunately it is not, in Bitcoin the decisions are mostly made by the miners.
In contrast, Avalanche separates the consensus problem from the governance problem. Each one is tackled modularly and independently. Consensus is achieved via a smart combination of gossip protocols with recurrent subsampling. While, fair governance is achieved via sampling and staking resulting in a direct tie between users and their decisions.
Disclaimer: Information in this article is for educational purposes only and cannot be taken as investment advice.