Why is the road to decentralization of L2 sequencers "difficult to get to the sky"?

Written by: Haotian

As explained in previous articles, both OP-Rollup and ZK-Rollup are stagnant on the issue of decentralized sequencer, does this mean that decentralized sequencer is completely unsolvable?

With this question in mind, I studied the next @MetisDAO and claimed to have built the first decentralized sequencer Layer2 project. It's not hard to discover the logic of Sequencer's decentralized technology, but it's "decentralization".

Sequencer solutions that adopt "hard distributed" technical consensus and "soft distributed" social consensus seem to be Make Sense, but there will be a completely different End Game!

On the ETH main network, users submit transactions to Mempool, which are collected and broadcast by Searcher, and finally packaged by miner Builder according to GAS high and low. This means that the transactions you submit may be MEV, and you need to follow the Nounce queuing mechanism, and the gas you pay is the cost of the EVM to perform the operation, resulting in expensive, slow, and easy to get caught in the transaction on mainnet.

By analogy, Sequencer in Layer 2 is a closed mempool where users submit transactions to Sequencer, and Sequencer is responsible for sorting and batching transactions into a block and submitting them to Mainnet. As a result, the ordering of layer 2 transactions is entirely up to Sequencer, and there is no such thing as nounce value queuing, and the gas paid is the average cost of the same batch of transactions.

In this way, layer2 transactions are fast and cheap, and in theory, Sequencer will fairly sort transactions according to their nonance, time of receipt, gas price, or random algorithm. However, due to the large power of Sequencer, there is a question mark over whether it will engage in MEV and whether malicious transactions will be inserted?

The current mainstream solution in the rollup market is a solution that defaults to Sequencer being centralized but worthy of "optimism".

On the one hand, because Sequencer is a core component of layer2, the potential risk of failure if distributed Sequencer is used at the beginning will affect the user experience, so most project parties tend to centralized operation and maintenance in the early stage to ensure security and stability.

On the other hand, when the project grows to a certain level and handles a large number of transaction needs, it uses supreme power, not only to control the gas pricing power but also to directly pump profits, at this time the market has grown to a certain size, and it is reluctant to implement decentralization and cede rights. This is the reason why the decentralization process of the layer2 Four Heavenly Kings Sequencer is slow.

However, they all chose Stack's strategic architecture. This is a strategic solution to open source the core technical code and try to share the key technical components.

By sharing the core Sequencer, creating a Superchain multi-chain combination architecture, and then co-governing through MPC multi-signature, governance right voting, etc., a socialized "transparent and decentralized" consensus has been reached.

This kind of softly distributed Sequencer solution is based on lofty ground, suitable for the premise that the market is large to a certain scale, everyone has absolute trust in the power center, and is desensitized to some trust friction problems (evil, challenges), such as the OP Stack.

What about the decentralized Sequencer solution provided by @MetisDAO?

To put it simply, it is a "hard distributed" solution for building distributed Sequencer nodes.

Multiple Sequencer nodes in layer2 will form a sequencer pool, and the sequencer will only be eligible to have block generation rights if it stakes 20,000 Metis tokens, and users can also select the corresponding Sequencer nodes in the form of staking, so that Sequencer block mining will receive a large number of token incentives, and user staking can also share the reward.

In addition, in order to prevent misbehavior in Sequencer, Metis has introduced L2 Rangers, which are used to sample blocks and verify them against their original state roots, including whether the order of transactions has been reversed, whether there are malicious transaction insertions, etc. Once Sequencer is found to be evil, the system will slash the assets pledged by the cheating node, and the validator can also get a share of the confiscated assets.

Sequencer joins the POS stake queue if it wants to be rewarded, and Verifier mining also strengthens the sampling verification work, through which the incentive and punishment model is strung together into a decentralized Sequencer system of work.

This "hard distributed" architecture, in which Sequencer is run by multiple agents and has a transparent reward and punishment mechanism, is also an effective way to constrain the rights of individual Sequencers and ensure fair ordering.

Above

Both soft and hard distribution are essentially a strategic tool.

The Stack open-source framework is more suitable for the expansion of Monolithic's large subject territory, which is more dependent on the entity's market position, brand prestige and other intangible trust assets, while the POS decentralized Sequencer is relatively more suitable for modular small entities to overtake in corners, after all, the most powerful consensus precipitated by the blockchain is the consensus model of node mining technology.

In the layer2 market situation of Rollup As A Service, the Stack framework will reduce the cost of developing layer2 for the project team, but the accumulated "trust" problem should be settled by all participants.

However, the direct hard distributed technical consensus has a clearer boundary on the issue of Sequencer, which is more suitable for a more eye-catching starting point and development expectations in the context of the slow progress of the overall layer2 technology and the cross-cutting of public opinion.

Of course, I've also looked at other decentralized Sequencer technical consensus schemes such as @EspressoSys, @AstriaOrg, @radius_xyz, etc.

Similar to Metis's logic, a single sequencer is designed as a multi-agent, but when it comes to this part of off-chain consensus, Esprsso uses Eigenlayer to extend the verification capabilities of ETH validators based on restaking.

Astria takes a more modular approach, providing layer2 developers with a fast modular Sequencer solution.

Radius encrypts all transactions in the Sequencer pool, eliminating the risk of misconduct and MEV. (Looking for an opportunity to carry out analysis)

In short, my opinion: the decentralization problem of Sequencer will not stop at the stack strategy of the four kings of Layer2, and although the decentralization scheme of technical consensus does have the risk of consensus overload, it will be more promising than the MPC multi-signature governance social consensus.