Ethereum's Rollup is centralized: Exploring Decentralization of Sequencers

1. Key Points

• Transaction ordering has become an increasingly serious issue in the second layer (L2) domain. L2 rollups primarily provide users with a low-cost trading venue, and then submit the transaction data to Ethereum and other first layers (L1).

• The sorter is responsible for grouping transactions in order. It receives users' unordered transactions, processes them off-chain into groups, generates compressed ordered transactions, and then sends them to L1.

• Rollups do not actually require a sequencer; this is simply a design choice to provide a better experience for users. However, currently all major L2 rollup projects have found that running a centralized sequencer is more convenient and cost-effective.

• The sorter controls the order of transactions, thus having the power to review transactions. They can also extract the maximum extractable value (MEV), which may cause economic losses to users. Furthermore, if the only centralized sorter goes offline, the entire rollup will be affected.

• The solution is a shared, decentralized sorter. This can not only solve the problems of censorship, MEV extraction, and availability but also enable cross-rollup functionality. Projects like Espresso, Astria, and Radius are developing innovative shared sorting solutions.

2. Introduction

With the popularity of the Ethereum L2 rollup ecosystem, the often-overlooked role of the sequencer is becoming increasingly important. The sequencer is responsible for transaction ordering, which can provide a better user experience, lower fees, and faster transaction confirmations. However, currently all major Ethereum L2 projects utilize centralized sequencers, which may be seen as not aligning with the spirit of cryptocurrency.

Although most companies have included decentralized sequencers in their roadmaps, there is still no real consensus on how to achieve this. It is worth noting that Arbitrum and Optimism have not made substantial progress in decentralized sequencers since the second half of 2021.

This report will delve into the role of sequencers and the current state of the Ethereum rollup space, exploring projects that are developing solutions, namely decentralized shared sorting networks. We will provide a detailed overview of these projects and the uniqueness of their solutions, and contemplate what this might mean for the future development of the Ethereum L2 rollup field.

3. What is a sorter?

Blockchain is essentially a distributed ledger consisting of timestamped transaction data sorted into blocks. This transaction data is initially unordered and can only be organized into blocks and executed after sorting, creating a new state of the blockchain. For L1 blockchains like Ethereum, transaction sorting takes place at the base layer itself.

In the most popular scalability solution on Ethereum, L2 rollup, transaction ordering has become an increasingly serious issue. L2 rollup provides users with an execution layer, and then submits the transaction data to L1. A batch of transactions submitted to L1 usually contains hundreds or thousands of compressed L2 transactions, thereby reducing the cost of sending data to L1.

In L2 rollups, the sequencer is responsible for ordering transactions into batches. It receives unordered transactions from users, processes them off-chain into batches, generates compressed ordered transactions, which can then be placed into blocks and sent to L1. The sequencer also provides users with near-instant receipts as "soft confirmations." A "hard confirmation" is received after the transaction is sent to L1.

Why do Rollups need to use sequencers, and why is it a problem?

The fundamental goal of the sorter is to improve the user experience. Using a sorter for L2 transactions is similar to a "fast lane", meaning lower fees and faster transaction confirmations. The sorter can compress hundreds or thousands of L2 transactions into a single L1 transaction, thereby saving gas fees. In addition, the soft confirmations provided by the sorter enable rollup transactions to be confirmed quickly.

Importantly, rollups do not require a sequencer; this is just a design choice made for better user experience. For instance, rollups can also use Ethereum L1 for sequencing. However, the Ethereum base layer can be relatively inefficient and expensive, especially considering the high volume of L2 transactions. This means that so far, every major L2 project has found it more convenient, cheaper, and easier for users to operate centralized sequencers.

Since the sorter controls the order of transactions, it theoretically has the right not to include user transactions in it. The sorter can also extract MEV from the transaction pool, which can cause economic losses to the user community. If there is only one sorter, the risk of centralization increases. In this case, if the sole sorter fails, the entire rollup will be affected.

With this setup, the sorter can be seen as a semi-trusted party for the user. Although the sorter cannot prevent users from using L2, it can delay the users' transactions, causing users to pay additional gas fees and extract value from the users' transactions.

The relevance of MEV

MEV refers to the value extracted from block production that exceeds the regular block rewards and gas fees. It is the value extracted by manipulating the order of transactions within a block. Common forms include front-running and sandwich attacks.

Given the role of the sequencers in L2 rollup, they can understand all user transactions off-chain. As these sequencers are often operated by the project itself or affiliated teams, many users are concerned about the inability to see potential MEV extraction. Even without these concerns, centralized sequencers can affect the trustlessness and decentralization of the protocol.

Current Status of the Sorter Market

Currently, all major Ethereum L2 projects rely on centralized sequencers. As more and more Ethereum transactions move to L2, despite the decentralized nature of Ethereum's validator set, a large number of transactions seem to be influenced by centralized forces in the form of a single sequencer.

As expected, most of these companies have included decentralized sorting in their roadmap. However, we should note that Arbitrum and Optimism launched their own solutions at the end of 2021, but have not made substantial progress in decentralized sorting.

Most top companies seem to be allocating resources to improve core products and features, rather than focusing on Decentralization. This is somewhat understandable, as in a highly competitive environment, focusing on Decentralization before having a competitive product is not in the best interest of any company. However, as network companies mature, this perspective is changing, and discussions are rapidly shifting towards sorting Decentralization and increasing credibility.

Other issues

There has been some discussion about the level of risk associated with relying on centralized sorters.

Although the sorter controls the order of transactions and can exclude user transactions and extract MEV, they ultimately cannot completely exclude users from rollup transactions. Users can bypass the sorter and submit transactions directly to L1 ( as long as they are willing to pay the increased gas cost ). While misbehaving sorters may lead to transaction delays and users incurring additional fees, they cannot ultimately completely censor. This may be one reason why some large L2 companies are less focused on decentralized sorters.

Perhaps the bigger issue lies in real-time performance. Given that the main rollup programs operate a single centralized sequencer, if these sequencers encounter issues, the entire rollup program will be adversely affected. Although users can still complete transactions by directly accessing L1, this is not a particularly sustainable method and is unlikely to work for most transactions.

4. Solution: Decentralization Shared Sorter

Overview

The new solution to the above problem is a decentralized shared sequencer. Although the solutions of different projects vary, the basic concept is the same. "Shared" refers to multiple different rollups using the same network. "Decentralization" refers to the concept of leader rotation, meaning that it is not always a single actor that sorts all transactions, but rather a leader selected from a group of decentralized actors. This helps prevent censorship and provides validity assurance.

The shared sequencer is designed to mitigate the MEV extraction problem, provide censorship resistance, and enhance the validity guarantees of rollups. Additionally, there are two other points worth noting:

1. Decentralization as a Service: The shared sequencer solution provides sequencing decentralization services for an arbitrary number of rollups. All these rollups will benefit from the censorship resistance and real-time capabilities that a decentralized network can offer without having to establish the network themselves.

2. Cross-rollup composability: Since these shared sequencer solutions are designed to handle transaction ordering for multiple rollups, they are able to provide unique interoperability guarantees that are currently unavailable. For example, users should be able to specify that a transaction on Rollup 1 can only be included in the block if different transactions on Rollup 2 are also included in the same block.

Many projects are researching shared sorting solutions. We will highlight several of them and their strategies in the following text.

Espresso

Espresso Systems is committed to building tools that bring Web3 to the mainstream, with a particular focus on L2 rollups and the Ethereum ecosystem.

The Espresso sorter is a decentralized shared sorting network designed for decentralized rollups, while providing secure, high throughput, and low latency transaction orders and data availability. Its design is independent of the virtual machine and can be used for non-Ethereum virtual machines, as well as zk virtual machines and optimistic virtual machines.

The core of the sorter is the consensus protocol HotShot. HotShot is based on the HotStuff consensus protocol and combines the latest developments from multiple different fields. HotShot is open and permissionless, decentralizing the power within the sorter network while ensuring security and efficiency, all while providing high throughput and fast final results.

Espresso Systems aims to achieve Ethereum-level security for its sequencer by leveraging the existing set of validators on Ethereum. There are two key reasons for this setup:

1. Security: By using the same validators as Ethereum, the sorter can achieve a level of security, effectiveness, and decentralization that is difficult to attain on its own.

2. Incentive Alignment: Conceptually, it makes sense to involve Ethereum L1 validators in the operation of the protocols that Ethereum L2 rollups run.

Espresso will seek to establish this partnership by resetting the contract, particularly with EigenLayer. Through EigenLayer repricing, users can stake their Ether and Ethereum liquid staking tokens across multiple protocols, thereby extending economic security beyond Ethereum itself.

Espresso also utilizes its efficient Tiramisu data availability solution to address the issues of limited block space and high transaction fees. Tiramisu has three layers: Savoiardi, Mascarpone, and Cocoa, which together ensure that data is provided to parties in need of it.

Espresso Systems considered flexibility and modularity when designing its protocol, using its sorter rollup device which can utilize any other data availability solution if Tiramisu is not desired.

Espresso has announced multiple partnerships, including EigenLayer, Polygon zkEVM, Injective, AltLayer, Caldera, and Spire.

Astria

Astria is building a shared sequencer network and is one of the leading companies to eliminate centralized sequencers. At the same time, they are also developing Astria EVM, which will be the first rollup supported by its shared sequencer network.

The shared sequencer network of Astria allows multiple different