Rollup's next step: how to implement a decentralized sequencer?
Author: Alex Beckett
Compilation of the original text: The Way of DeFi
Author: Alex Beckett
Compilation of the original text: The Way of DeFi
A centralized sequencer (sequencer) is the current common way for rollup, and most projects (if not all) plan to implement a decentralized sequencer in the future. There is no right way to decentralize, as there are multiple options for different outcomes.
Permissionless PoS with leader election
To instantiate a set of permissionless sequencers in place of a centralized sequencer, the minimum requirements are some form of Sybil resistance and a leader election mechanism. Sybil resistance is to impose an execution cost on malicious behavior. Using Proof of Stake (PoS), the cost can be imposed on the sequencer by missing rewards (such as inactivity leaks), or by directly reducing the stake (slashing).
Since rollup uses the L1 (we call it the base layer) for consensus, all that is needed is a leader election mechanism to rotate the sequencers. Nodes get ordering privileges relative to the percentage of stake they have (say a 10% stake will yield ordering duties about 10% of the time). There are several leader election algorithms bundled with the consensus protocol that work well as stand-alone mechanisms for sequencer sets.
The staking mechanism can be established at the base layer or directly in rollup through smart contracts. The point is that, like any monolithic chain, individuals only need to join the sequencer set by staking native tokens and meeting minimum hardware requirements.Permissionless PoS approach with MEV auctionSequencer responsibilities can be awarded through auctions instead of the designated leader election mechanism. The auction is designated as
MEV Auction (MEVA)
, as the entity with the highest bid has the right to become a sequencer and extract any MEV during the specified sequencer period. The period can be of any length, although there is a trade-off between frequent auctions versus infrequent auctions.
MEV auctions may sound attractive because sequencing rights are permissionless. Anyone can bid to become a sequencer, even if only the highest bidder gets that right. Awarding the sequencer role to the highest bidder ensures that only the best-funded bidder wins each time. Additionally, the sequencer that is able to extract the most MEV will be able to grow its capital base faster than the other sequencers. Finally, MEVA centralizes and incentivizes order monopoly. Monopoly stems from the high entry barriers to becoming a sequencer and winning MEVA, and the pricing power a sequencer can exert.
While winning MEVA doesn't give the sequencers control of the system, they can still be slashed.
Rollups do generate income from MEVA, which can be used for various purposes, such as protocol making or public goods funding. However, these revenues do not come for free, MEVA generates revenue at the expense of users affected by MEV withdrawals. Since MEV extraction is essentially an invisible tax on users, MEVA is simply a redistribution of wealth from users to MEVA income beneficiaries.
Paying some type of tax for things like development and financing public goods is arguably a noble cause, and one I support. Gaining revenue from taxes in the form of MEVA is far from an ideal solution. There are much better and more thoughtful ways to implement taxes.
Permissive sequencer set with fair ordering
If the goal is to lower MEV, a slightly different approach can be taken. The sequencer role can be distributed among a trusted/semi-trusted set of participants. Each sequencer participates in a consensus process to determine transaction ordering (also known as fair ordering). Through consensus, the concept of "fair" ordering can emerge, similar to first come, first served.
A user makes a transaction, which is broadcast to all sequencers;
Sequencers build lists of transactions in the order they claim they arrived;
Merge all sequencer lists using a fair sorting algorithm to create some form of consensus view of first-come, first-served order;
The final ranking after the consensus process is released to the base layer;
It is worth noting that the consensus process introduces an honest majority assumption. If most people are honest, then first-come, first-served ordering will be supported despite attempts by other malicious sequencers to manipulate the ordering. If most sequencers are dishonest, the ordering can be manipulated in arbitrary ways to extract MEV.
Since sequencer sets are permissioned, active governance is required to allow sequencers into the set and provide recourse for handling malicious or dishonest sequencer behavior. The extent to which this process involves the community is unclear, although it can be difficult to discern order manipulation from network issues that lead to it being manipulated. Discerning order manipulation is difficult, which is why slashing doesn't necessarily penalize the sequencer programmatically.
in conclusion
in conclusion
