The New PoS Consensus for Ethereum
Post-merge With Gasper, Ethereum is switching to a new consensus algorithm. To reach consensus, Gasper mixes LMD GHOST and Casper FFG (how to determine finality and fork-choice rules, respectively). Gasper is a consensus mechanism that prioritizes liveliness over safety, much like its previous Nakamoto-style consensus. Consensus algorithms that put safety first fail if they don’t get the necessary number of votes. As a result, the chain breaks (the opposite of “liveness”). If rogue actors have tampered with the bulk of the hash power/votes, chains that favor liveness will still produce blocks according to their “heaviest/longest chain” rule set.
The consensus protocol that uses PoS on Beacon Chain has a name. It is Casper the Friendly Finality Gadget. The PoS and BFT consensus models work together. Combined in the partial consensus algorithm known as Casper. Although the Nakamoto Consensus and Casper’s fundamental design were both based on the PBFT consensus mechanism, they both still identify the “real” chain as the one with the most attestations. Casper FFG decides which blocks are included in the chain.
After receiving enough votes from its 440,000+ validators, Casper approaches finality through recurrent checkpoints. Epochs and slots are two different categories of checkpoints. It adds a block to the beacon chain during a slot, which happens predictably every 12 seconds in the PoS system. Each slot has a distinct validator assigned to propose a new block in advance. A group of 32 slots (6.4 min) is called an epoch. Additionally, a committee made up of at least 128 validators is assigned to each slot.
Ethereum Post Merger
Each proposed block is examined by committees, and who vouch for its legitimacy. The committee members broadcast a cryptographic attestation of the block to the other validators and network nodes after successfully validating the block. Finally the fork-choice rule LMD GHOST determines the current head of the chain based on the validator attestations.
Epochs taken into account for chain finalization have further trimerous effects. And for the periodic checkpoints as well. If more than 66% of the validators attest throughout an epoch, it seems justifiable. The first epoch will be finalized and the transactions will become irreversible if a second epoch with > 66% follows the first.
According to the protocol’s “correlation penalty,” validators lose their ETH incentives if they decline to participate when requested and see their stake “slashed” if they submit multiple votes for the same block or make contradictory proposals. The amount of ETH taken depends on how many dishonest validators the protocol eliminates at once.
If a validator punished on its own. This punishment can result in a slashing of approximately 1% of their stake. It can result in a collective down fall of close to 100%. The penalty, made by the protocol halfway through a forced exit period. That starts on day 1 with an instantaneous penalty. And finally moves on to day 18 with a correlation penalty. And ends on day 36 with ejection from the network. Additionally, because the dishonest node is present on the network but isn’t casting votes, it suffers a daily minor attestation penalty.
