Analyzing the New Version of Metis: The Decentralization of the Lowest Layer 2 Gas in Progress | CatcherVC Research

Text / SA, CatcherVC

Key points of this article: · The fundamental issue with OP Rollups like Optimism and Arbitrum is the centralization of Sequencer nodes, which requires reliable solutions.

· Metis attempts to be the first to achieve decentralization of Sequencers, opening a Peer Node network and transferring the power to run Sequencer nodes to community members or other organizations.

· Metis has changed the storage layer structure, altering the way data is published on Ethereum. By integrating Memolabs, it significantly reduces storage costs, becoming the one with the lowest Gas fees among mainstream Layer 2s.

· By introducing new mechanisms, the new version of Metis after integrating Memolabs still maintains reliable security and data availability. The team has assessed potential situations and formulated effective measures.

· Metis supports node operators to register as DACs (institutional-level DAOs) to obtain continuous token rewards; at the same time, it provides an easy one-stop DAO hosting service to reduce the operational difficulty of DAOs and opens up Community Ecosystem Governance (CEG), further transferring the maintenance rights of the Layer 2 network to community members.

Problems with traditional OP Rollups

With concepts like Web3, the Metaverse, and NFTs entering the public eye over the past year, the Crypto industry has officially entered a period of rapid growth. With the pursuit of various capital and massive users, Ethereum, leveraging its first-mover advantage, has become the core of the entire Web3 narrative. Its system architecture has fully realized decentralization and security after a long evolution, becoming a true "stellar public chain." However, the inefficiency severely limits the development of this public chain. Compared to VISA, which can handle thousands of transactions per second, Ethereum's TPS of less than 20 feels like an antique from a bygone era, far from Vitalik's grand vision of a "world-class decentralized application platform."

To meet the enormous demand of the Web3 market, various solutions such as sidechains, new public chains, and Rollups have taken the historical stage one after another. Star projects like BSC, Polygon, Solana, Arbitrum, and Optimism have divided the traffic, but their inherent flaws have become increasingly clear. Due to TPS being constrained by block generation speed, major Layer 2s or new public chains have almost all compressed the number of nodes or decoupled "consensus" from the block generation process, directly reducing block generation time but severely weakening decentralization and system security.

Taking Optimism as an example, it uses a single miner node called Sequencer to generate blocks in Layer 2 in seconds. New blocks do not need to be immediately handed over to other nodes for verification and can be finalized locally, saving a lot of time; at the same time, since there is only one block generation node, the allocation of "accounting rights" is determined, allowing the direct elimination of the POW process (the step of randomly allocating accounting rights).

（Core architecture of Optimism）

By shortening the block generation process, the local blocks in Layer 2 can be finalized in as little as 1 second or even less. After a user initiates a transaction request, they can receive results in as little as two or three seconds, comparable to WeChat Pay.

However, at this point, the new blocks in Layer 2 have not been audited by verification nodes, which poses the possibility of non-compliance. In response, the Sequencer must regularly publish local block copies on Layer 1, including transaction data and state roots (which are associated with account information on Layer 2). The Verifier nodes in Layer 2 will automatically read the content published by the Sequencer and audit it to determine whether the Sequencer is suspected of fraud.

Essentially, Optimism treats Ethereum as a "court" for disclosing data and resolving disputes, with the key point being how often the Sequencer publishes data on Layer 1. If the Sequencer takes a long time to submit local data, it will undoubtedly delay the auditing progress of the Verifier, and it will take a long time to reach consensus among nodes, which will severely undermine the reliability of Layer 2.

According to Optimism's official browser, the frequency at which the Sequencer publishes state information on Ethereum can be as slow as once every 30 minutes, meaning that after the Sequencer generates a block, the Verifier has to wait half an hour to conduct an audit. In contrast, new blocks on Ethereum can be audited by the entire network of nodes in just 13 seconds. Clearly, there is a serious asymmetry of information between Optimism's Sequencer and Verifier nodes, and the reliability of the consensus mechanism is far lower than that of Ethereum.

【Readers can refer to CatcherVC's previous article: 《In-depth Analysis of Optimism: Basic Architecture, Gas Mechanism, and Challenges》】

In response, Arbitrum, which is also part of the OP Rollup faction, has shortened the interval for submitting state information to 2-5 minutes, allowing Verifier nodes to conduct state audits as quickly as possible, significantly reducing the information gap.

However, Arbitrum has the same flaw as Optimism: the Sequencer node responsible for block generation is operated by the official team, and it has not transferred "accounting rights" to the outside world. The reason is that the Sequencer holds too much real power, which relates to the asset security of all users. In the absence of a well-designed mechanism, it cannot guarantee "procedural justice." To be safe, both Arbitrum and Optimism's block generation nodes are backed by the official team to compensate for the current system's imperfections.

The consequences of this approach are clear: Arbitrum and Optimism effectively become centralized operators. Although both allow users to freely run Verifier nodes and challenge the Sequencer, the official team still has absolute authority over the appointment and dismissal of the Sequencer. Thus, even if a Verifier points out that the current Sequencer is acting maliciously and forces it to step down, the new Sequencer will still be appointed by the official team.

Essentially, the block generation power of Layer 2 is concentrated in the hands of the official teams of Arbitrum and Optimism, relying on "credit" rather than "procedural justice." At the same time, having the official team run the Sequencer nodes brings another major issue: the small number of block generation nodes and their concentrated physical locations make them vulnerable to DDoS attacks or other types of single points of failure.

Taking Arbitrum as an example, its Sequencer node has gone down twice, attracting widespread attention. On September 14, 2021, both Arbitrum and Solana went down due to DDoS attacks, as the block generation nodes received too many transaction requests in a very short time, ultimately leading to a crash; on January 10, 2022, Arbitrum's Sequencer node went down again, with the official team stating that the node experienced hardware failure, and the backup node equipment did not complete the handover in time, ultimately causing a "single point of failure" that halted the entire Arbitrum network.

It is easy to see that the flaw of centralized systems like Arbitrum and Optimism lies in the excessive concentration of resources, with only a few or a single node responsible for block generation, which puts them under significant access traffic and is prone to single point failures; at the same time, the official monopoly on block generation power makes "fraud proofs" and "challenge mechanisms" virtually useless, failing to fundamentally curb the issue of node malfeasance.

Regarding their inherent flaws, the officials of Arbitrum and Optimism have stated that they will gradually improve and implement decentralization in the future. However, at present, neither has provided reliable solutions, and the specific implementation of decentralization is still far off.

In adherence to the principles of decentralization, Metis, also an OP Rollup solution, has recently officially begun to reform its system architecture, attempting to achieve decentralization in both architecture and economics for Layer 2.

· By opening a Peer Node network, Metis transfers the power to run Sequencer block generation nodes to community members or other organizations, promoting rapid information synchronization between Sequencers and other peer nodes to prevent malfeasance;

· Metis supports node operators to register as DACs (institutional-level DAOs) to obtain continuous token rewards.

· Metis has officially opened Community Ecosystem Governance (CEG), further transferring the rights to maintain the Layer 2 network ecosystem to community members.

Through these methods, Metis plans to be the first to achieve decentralization of Layer 2.

In addition, Metis has changed the way data is backed up on Ethereum. In the Peer Node network, which can immediately verify the local blocks of the Sequencer and prevent malfeasance within the Layer 2 network, Metis backs up transaction instructions to the off-chain decentralized platform Memolabs, providing the storage location of transaction data on Memolabs on Layer 1, while the StateRoot corresponding to each transaction is still published on Layer 1.

In response to potential "challenges" and "fraud proof" scenarios, Metis has added other functions, allowing challengers to restore the original data of each transaction instruction on Layer 1 when such scenarios occur, enabling "fraud proofs" to be completed without hindrance, making the existing version and the old version's mechanisms equivalent.

（Core architecture of the new version of Metis）

By introducing peer nodes and integrating the Memolabs storage layer, Metis has transformed the storage tasks originally handled by Ethereum into a shared responsibility among peer nodes, Ethereum, and Memolabs, while introducing new mechanisms to ensure reliability. With the burden of storage shared by the other two, Metis can judiciously reduce the amount of data published on Ethereum, thereby lowering Gas consumption and significantly reducing Layer 2 transaction fees.

In the following text, the author will interpret Metis's implementation of the Peer Node network and integration of Memolabs storage and other important measures.

Peer Nodes: Achieving Rotation of Block Producers ------ Sequencers

In traditional OP Rollup solutions like Optimism and Arbitrum, the block generation node is uniquely determined: only one Sequencer executes transactions and packages blocks. This directly eliminates the randomness of block generation nodes, and at the beginning of each block generation cycle, the system no longer wastes time selecting a block producer—by contrast, Ethereum must randomly select a block generation node through the POW or POS process (after merging) before each new block is generated, severely delaying time.

However, the randomness of block generation nodes can significantly reduce the probability of single-point malfeasance. Since accounting nodes are frequently rotated, the likelihood of a malicious node controlling the legitimate ledger is minimized. Even if a malicious node gains accounting rights for a new block, if the block it publishes is non-compliant, it will still be rejected by other honest nodes. Ultimately, honest nodes will re-elect a new block producer and republish a compliant block, effectively sidelining the malicious node.

In this case, theoretically, as long as 2/3 of the nodes in the network are honest, they can effectively constrain malicious nodes, which is the famous PBFT mechanism (Practical Byzantine Fault Tolerance); currently, the fault tolerance rate of Bitcoin and Ethereum nodes can even reach 1/2. However, the effectiveness of such fault tolerance algorithms is based on having a sufficient number of nodes. When there are a large number of nodes, it is difficult for malicious nodes to recruit a significant number of nodes and form a conspiracy, which is when PBFT becomes effective. When the number of nodes participating in block generation is small, PBFT is no longer applicable, and the likelihood of single-point malfeasance becomes very high.

Existing OP Rollups, including Optimism and Arbitrum, almost all assume that the Sequencer block generation node will not act maliciously. If the Sequencer does engage in malfeasance, Verifier nodes are allowed to "impeach" it, a process referred to as "challenging." However, the problem is that there is not an immediate data synchronization between the Verifier nodes and the Sequencer, leading to a delay.

As mentioned earlier, the data synchronization delay for Optimism nodes can exceed 30 minutes, meaning that after the Sequencer generates a new block, it takes half an hour for the verification nodes to audit it, which poses potential security risks. Although Arbitrum has reduced the delay to a few minutes, it has not opened the operation of the Sequencer to organizations outside the official team, which is not conducive to economic decentralization and relies on the project's "credit," severely violating the principle of "procedural justice" in blockchain.

Moreover, since neither Optimism nor Arbitrum has issued tokens, they cannot provide strong incentives for Verifier node operators, which is not conducive to expanding the number of nodes, making Layer 2 more like a consortium chain rather than a public chain.

To avoid the aforementioned issues, Metis has made significant improvements to the original architecture of Optimism, with the most important measure being the opening of Peer Nodes (Peer Node).

· Traditional public chains like Bitcoin and Ethereum are composed of a P2P network of peer nodes, which frequently synchronize information among themselves to ensure state consistency; at the same time, each peer node can voluntarily become a miner and participate in block generation. After a new block is generated, it will be propagated to other peer nodes for auditing.

· Metis has built a Peer Node network called Sequencer Pool, allowing community members to run peer nodes, which act as Sequencers in a rotating manner, achieving a "turn-taking" system to solve the centralization problem of Sequencer nodes in OP Rollups;

（Peer Node network）

· Currently, after the Sequencer generates a block, it will synchronize the new block to other peer nodes for auditing to prevent single-point malfeasance. After a certain period, the Sequencer will change, thus achieving the decentralization of accounting rights.

· Each block generation cycle of a typical public chain involves a random selection of the block producer, which wastes a lot of time. Currently, the number of peer nodes in the Sequencer Pool is not as large as that of major public chains, and the rotation cycle of block producers is relatively long. In each cycle, the Sequencer remains singular. In the future, Metis will gradually shorten the rotation cycle and introduce a new timestamp generation mechanism.

· Metis supports community members to run peer nodes and provides token incentives for them. Peer node operators often register under the name of DAC (institutional-level DAO), with hardware requirements of at least an 8-core CPU, 32GB of memory, and a certain amount of Metis tokens to be staked.

Essentially, the Sequencer Pool, originally a subnet under the Metis network, has become a "committee" composed of peer nodes, whose function is to act as or supervise the Sequencer, and in form, it reflects the shadow of POS public chains.

According to the plan being implemented by Metis, the Sequencer Pool has started operating with a scale of a dozen peer nodes. At this network scale, the time complexity of communication among peer nodes is very low, allowing for immediate consensus on new blocks. At the same time, different peer nodes can act as network load, meeting external access requests, so users do not have to rely solely on data provided by a single node.

（Architectural vision from Metis white paper）

Currently, Metis effectively obtains two layers of security from the Peer Node network and Verifier nodes. Among them, peer nodes can verify the local data of the Sequencer in Layer 2 in real-time, while Verifiers are mainly responsible for verifying the data submitted by the Sequencer to Layer 1.

In the future, Metis plans to significantly expand the number of peer nodes in the Sequencer Pool to enhance security and include Verifier nodes in the Sequencer Pool list so that all peer nodes can serve as both Sequencers and Verifiers. At the same time, Metis plans to introduce new algorithms and timestamp generation mechanisms to ensure decentralization while maintaining high efficiency by "changing the Sequencer every few blocks."

New storage structure------"Entities should not be multiplied beyond necessity"

In most public chains or Layer 2s, the database that records user information adopts a tree structure, known as the state tree, with the hash value of the tree root referred to as the StateRoot. When a transaction instruction is executed, the state of certain accounts inevitably changes, and at this point, the hash value of the state tree root will also change. It can be said that the execution of each transaction generates a new StateRoot, and from a temporal perspective, the two are one-to-one corresponding.

If we list each transaction's instruction content and the corresponding StateRoot in chronological order, we can obtain an accurate ledger. In traditional OP Rollup solutions like Optimism, what the Sequencer stores on Ethereum is this content.

Verifiers will read this content and check its accuracy. Generally, Verifier nodes will execute the transaction instructions in chronological order, calculating a batch of StateRoots, and then the Verifier only needs to compare its calculated StateRoot with the StateRoot submitted by the Sequencer, which is akin to a teacher grading students' math homework without knowing the standard answers in advance, relying on mental calculations.

If a Verifier finds that a transaction instruction or the corresponding StateRoot submitted by the Sequencer has issues, it will initiate a "challenge" and provide a "fraud proof."

In Optimism and the old version of Metis, the Sequencer would publish transaction instructions and the corresponding StateRoots on Ethereum, effectively using Ethereum as a storage layer while also leveraging the Ethereum network to handle the "challenge" process. Although this ensures data availability, it incurs very high Gas costs.

Taking the batch of transactions published by Optimism on Ethereum as an example, this batch contains a total of 204 transaction instructions, consuming over $211 in Gas fees, which means the storage cost for a single transaction instruction exceeds $1; considering the Gas required to store the corresponding StateRoots for this batch of transactions, the storage cost for a single transaction on Optimism can reach $1.5, which is still too high for most users.

In response to this issue, Metis has made significant adjustments recently. Metis has eliminated the step of directly storing transaction instructions on Ethereum, transferring transaction batches to Memolabs, a platform similar to Filecoin but with lower storage costs and faster data retrieval speeds. By integrating the Memolabs storage layer, the Sequencer first stores a large batch of transaction instructions in Memolabs, then publishes the storage index of this transaction batch on Ethereum, allowing Verifier nodes to read the original transaction data from Memolabs using this index value.

（Comparison of the amount of data published by the old and new versions of Metis on Layer 1, showing a significant reduction in the new version）

At the same time, since the StateRoot is more important than transaction data, it is still stored on Ethereum.

In summary, this aligns with Occam's Razor principle: "Entities should not be multiplied beyond necessity." The philosophy of Metis is that content that does not need to be stored on Ethereum can be equivalently replaced through other means. This can save storage costs and reduce the cost pressure on users.

Through this storage structure, Metis can significantly compress storage costs, reducing the transaction fees for a single Layer 2 transaction to just a few cents. Currently, Metis has become the one with the lowest Gas fees among mainstream Layer 2s.

However, the above practices of Metis have given rise to other issues: will changing the storage structure affect security or data availability? In this regard, we will analyze various potential outcomes.

The security and data availability issues of Metis and OP Rollups have two aspects, the first being:

· When the Sequencer executes transactions in Layer 2, it immediately finalizes them locally, temporarily possessing "finality." The specific scenario is: after a user initiates a transaction request on the Metis network, they will receive results in a few seconds. The question here is whether the temporary "finality" granted by the Sequencer is reliable?

Since the Sequencer of Metis will immediately synchronize information to the peer nodes in the Sequencer Pool after generating a block, the nodes can promptly audit the block content. If they find that the Sequencer has submitted a non-compliant block, they can remove it from the Sequencer Pool. Therefore, the security here is comparable to that of ordinary public chains. At the same time, external parties can choose information sources among multiple peer nodes, eliminating the need to trust a single node, thus ensuring data availability.

The second issue is:
· After Metis transfers transaction data to Memolabs, will the verification process and challenge mechanism be affected? Will new nodes joining the Metis network encounter inconveniences when synchronizing historical data?

This involves multiple potential situations, which can be discussed categorically. Since Metis still publishes StateRoots on Ethereum, the availability of StateRoots will not be affected. The availability of transaction data pertains to Verifier nodes or new nodes joining the Metis network.

For the latter, new nodes only need to synchronize historical data through other Verifiers or peer nodes, and they can also read transaction data from Memolabs and StateRoot records from Ethereum. Currently, Metis has over 80 privately operated Verifier nodes, already possessing strong data availability. Considering that the number of Verifiers is still expanding, new nodes will not face significant issues when synchronizing historical data.

For existing Verifier nodes, the challenge lies in: whether they can smoothly obtain transaction data and check the corresponding StateRoots. If they find discrepancies in the content submitted by the Sequencer, can they successfully initiate a "challenge" on Ethereum?

For this issue, we can analyze the following scenarios separately:

1. If the Sequencer provides a Memolabs index on Ethereum, allowing the Verifier to successfully read the transaction data, and upon inspection, these transaction instructions are correct (e.g., digital signatures are valid), then the remaining check is whether the StateRoot stored on Layer 1 matches.
· If audited, each transaction can match its corresponding StateRoot, then the Verifier successfully completes data synchronization without needing to initiate a "challenge." There are no issues at this point.

· If the Verifier finds that a transaction instruction and StateRoot do not match, the StateRoot must be incorrect. The Verifier can request the Sequencer to disclose the transaction data corresponding to the erroneous StateRoot on Layer 1.
If the Sequencer agrees, the "challenge" process proceeds smoothly, and the Sequencer will be penalized;
If the Sequencer disagrees, the Verifier can write the transaction data it read from Memolabs onto Ethereum to complete the "challenge," and the Sequencer will similarly be penalized.

Clearly, in the above scenarios, data availability and the "challenge" mechanism remain unaffected.

2. If the Sequencer stores forged transaction instructions in Memolabs (with invalid digital signatures), the Verifier must initiate a "challenge"; additionally, the Verifier must obtain the correct Layer 2 native transaction instructions to verify the correctness of the StateRoot.

· In this case, the Verifier can request the Sequencer to publish the relevant transaction batch on Ethereum, which will incur substantial Gas costs for the Sequencer, effectively serving as a penalty;
· If the Sequencer refuses, the Verifier can disclose the erroneous data it read from Memolabs onto Layer 1, initiating a "challenge," and the Sequencer will face even harsher penalties.

Under normal circumstances, if the Sequencer is successfully challenged by the Verifier, the losses it incurs will far exceed the Gas costs of publishing the transaction batch on Layer 1. Therefore, if the Verifier requests the Sequencer to publish transaction data on Layer 1, it will certainly disclose the correct transaction data.

At this point, the Sequencer must publish the individual transaction batch required by the Verifier, containing hundreds or thousands of transaction data, and the Gas consumed when publishing on Layer 1 will be very high, potentially reaching thousands of dollars, effectively serving as a penalty.

From the above discussion, it can be concluded that data availability and the "challenge" process remain unaffected.

3. If the Sequencer publishes a false Memolabs storage index on Layer 1, preventing the Verifier from successfully reading the data contained in the transaction batch, it can request the Sequencer to disclose the transaction batch on Layer 1 as previously described. If it refuses, the Verifier can obtain the corresponding data from peer nodes and continue subsequent verification work or initiate a challenge.

For Sequencers that repeatedly publish forged transaction data, publish incorrect Memolabs storage indices, or refuse to cooperate with Verifier requests, members of the Metis community can mark the malicious Sequencer operators through DAO governance voting, revoking their permissions, similar to a political "impeachment."

Through these meticulously designed mechanisms, Metis can protect the rights of Verifier nodes. However, to prevent Verifiers from abusing their power and maliciously demanding Sequencers to write transaction data on Layer 1, attacking honest Sequencer operators through Gas consumption, Metis has made the following requirements:
· If a Verifier requests the Sequencer to write transaction data on Layer 1, they must stake a certain amount of funds in advance to obtain whitelist eligibility, and each time they issue such a command to the Sequencer, they must incur a fee; this fee has been carefully measured to prevent Verifiers from frequently making unreasonable demands on the Sequencer.
· Any node can initiate a "challenge" and "fraud proof," theoretically allowing these nodes to cooperate with each other to ensure data availability and security.

Summary
Based on the core arguments presented earlier and recent updates from Metis, the following conclusions can be drawn:

· The fundamental issue with OP Rollups like Optimism and Arbitrum is the centralization of Sequencer nodes, which requires reliable solutions; Metis attempts to be the first to achieve decentralization of Sequencers.

· Metis has opened a Peer Node network, transferring the power to run Sequencer block generation nodes to community members or other organizations, implementing a rotation system and promoting rapid information synchronization between Sequencers and other peer nodes to prevent malfeasance;

· Metis has changed the storage layer structure, altering the way data is backed up on Ethereum. By integrating Memolabs, Metis has significantly compressed storage costs and has become the one with the lowest Gas fees among mainstream Layer 2s.

· Through meticulous mechanism design, the new version of Metis after integrating Memolabs still possesses strong security and data availability. The Metis team has assessed potential situations and formulated corresponding measures;

· To further reduce the power of the Sequencer, Metis plans to add the role of Proposers in the future, responsible for submitting the StateRoots corresponding to each transaction to Ethereum after the Sequencer publishes transaction data, thus forming a stronger decentralized check and balance.

· Metis supports node operators to register as DACs (institutional-level DAOs) and provides them with continuous token rewards. In this regard, the issued token Metis has an advantage over the non-issued Optimism and Arbitrum;

· Metis provides an easy one-stop DAO hosting service, reducing the operational difficulty of DAOs and DAC organizations, and opens up Community Ecosystem Governance (CEG), further transferring the rights to maintain the Layer 2 network ecosystem to community members. Currently, the Metis ecosystem has 500 DAC organizations with nearly 5,000 members.

· After integrating the Memolabs decentralized storage layer, DAO organizations within the ecosystem can transfer non-public data to Memolabs, while the corresponding storage index can only be obtained by whitelisted users, ensuring that DAOs maintain their privacy.

· In the future, Metis will support a sharded structure of multiple subnets, allowing different DAO organizations to run MVM virtual machines with independent states to achieve a multi-chain sharding mechanism similar to ETH2.0.

· Recently, Metis has launched the NFT cross-chain bridge function, and with ultra-low Gas fees, Metis is committed to building the best platform for NFT users;

· In the future, under conditions of sufficiently strong system fault tolerance, Metis will judiciously shorten the challenge period, becoming the most convenient Layer 2 for cross-chain transactions.

References
The Tech Journey: Lower Gas Costs & Storage Layer on Metis

In-depth Analysis of Optimism: Basic Architecture, Gas Mechanism, and Challenges