(De)Centralization of Ethereum Rollups
Written by Jason Yu (Research Team Fall 2024)
Introduction
Ethereum, as one of the leading blockchain platforms, continues to face substantial challenges in scaling its network to accommodate the increasing demand for decentralized applications (dApps) and financial systems. To address these limitations, Ethereum has adopted a rollup-centric roadmap, wherein scalability is achieved through Layer 2 (L2) solutions while maintaining the decentralization and security of the Layer 1 (L1) blockchain. Rollups, which are primarily categorized into Optimistic Rollups and Zero-Knowledge (ZK) Rollups, have become integral to this strategy by enabling higher transaction throughput and reduced costs without compromising Ethereum’s core principles.
This paper seeks to critically examine the future trajectory of Ethereum rollups by analyzing their various stages of development and assessing their implications for the security and centralization of the Ethereum ecosystem. It will provide a balanced evaluation of both the advantages and limitations inherent in the current rollup roadmap, exploring whether the present direction is optimal or if alternative approaches should be considered. Particular attention will be given to how the sequencers impact Ethereum’s ability to address the blockchain trilemma of scalability, security, and decentralization.
The primary objective of this research is to contribute to a deeper understanding of Ethereum rollups and their role in shaping the network’s future. By offering insights into the current landscape of rollup technologies, this paper aims to foster informed discussions within the blockchain community regarding Ethereum’s scalability strategy. While targeted at readers with some foundational knowledge of blockchain and Ethereum, this paper aspires to provide a comprehensive analysis that enhances understanding and promotes further exploration of Ethereum’s evolving ecosystem.
Sequencers, Centralized
The issue of sequencer centralization in rollups is a significant concern that undermines the decentralization ethos of blockchain technology and introduces critical vulnerabilities. Sequencers are responsible for ordering, batching, and submitting transactions from L2 rollups to Ethereum’s L1. While centralized sequencers improve efficiency and throughput, they come with a host of risks, including censorship, single points of failure, and the potential for economic exploitation through Maximum Extractable Value (MEV).
Centralized sequencers have the power to selectively exclude transactions from being included in a batch. This undermines blockchain’s core principle of being open and permissionless. For example, in cases like the $2.6 million exploit on Linea, the centralized sequencer was paused to censor attacker addresses, sparking debates about the dangers of such control. While users can bypass sequencers by submitting transactions directly to L1, this is costly and impractical for most use cases.
The other issue is that if a centralized sequencer goes offline due to technical failures or malicious attacks, the entire rollup halts. This has already occurred in several instances where L2 block production was disrupted by software bugs or L1 gas fee outages. Although mechanisms like “force-include” or “escape hatches” exist to allow users to bypass sequencers and submit transactions directly to L1, these solutions often lead to instability and inefficiencies on the rollup.
Centralized sequencers have significant control over transaction ordering, enabling them to extract MEV by reordering or frontrunning transactions. This not only harms users economically but also creates an unfair advantage for sequencer operators. For example, private mempools on platforms like Optimism’s OP Mainnet exacerbate this issue by allowing sequencers to exploit transaction visibility.
One prominent example is the incident involving the Linea rollup, a Consensys-backed zkEVM solution, where users lost over $2.6 million due to an exploit on a decentralized exchange. In response, the Linea team paused its centralized sequencer and censored attacker addresses to protect users and builders in the ecosystem. While this action was intended to mitigate further losses, it has raised significant concerns about the power of centralized sequencers to unilaterally halt operations and censor transactions. Such actions contradict blockchain’s core principles of censorship resistance and decentralization, exposing the vulnerabilities of centralized control.
Why Centralization Persists
Despite these risks, most rollups continue to rely on centralized sequencers due to practical considerations like cost efficiency and development complexity. Running a centralized sequencer is cheaper and simpler compared to implementing decentralized alternatives. It allows rollups to offer low transaction fees and high throughput, which are critical for user adoption.
Decentralizing sequencers requires significant technical innovation and coordination among stakeholders. Many rollups have included decentralization in their roadmaps but have yet to make substantial progress due to the complexities involved. They also ensure low latency and consistent transaction processing, which are essential for maintaining user experience. Decentralized models may increase latency, making them less attractive in the short term.
Sequencers, Decentralized
There are a few different kinds of solutions that could potentially solve the centralized sequencer problem, one of them being using a shared sequencing network. This aims to provide decentralization-as-a-service by allowing multiple rollups to share a common sequencing layer. This approach enhances censorship resistance, liveness guarantees (low/no downtime), and interoperability between rollups. For example, shared sequencers enable atomic cross-rollup composability while allowing rollups to maintain sovereignty by storing transaction data on L1.
Another one could be the decentralized networks distribute sequencing responsibilities across multiple nodes, reducing the risk of censorship and single points of failure. Anyone can participate in sequencing by running a node, creating a fairer system that aligns with blockchain principles.
Based rollups, also known as L1-sequenced rollups, are gaining traction as a solution that leverages Ethereum itself as the sequencing layer. Unlike traditional rollups with dedicated sequencers, based rollups rely on Ethereum’s validators for transaction sequencing, inheriting its security, decentralization, and liveness guarantees. This eliminates the risks of centralized sequencers, such as censorship and single points of failure, while ensuring seamless interoperability across rollups by aligning transaction sequencing with Ethereum’s block structure.
The Challenge with Data Availability
Data availability (DA) is a critical aspect of rollup security and functionality, ensuring that all necessary data for validating off-chain transactions is accessible to participants. Rollups, such as Optimistic Rollups and ZK Rollups, execute transactions off-chain and periodically post summaries of transaction data to Ethereum’s L1. While this approach enhances scalability and reduces congestion on Ethereum, ensuring robust data availability presents significant challenges that can compromise the security, decentralization, and efficiency of rollup systems.
Data availability ensures that all transaction data required to reconstruct the rollup’s state is accessible to validators, users, and nodes. For example, Optimistic Rollups rely on fraud-proof mechanisms to dispute invalid state transitions. Validators must access transaction data posted on L1 to construct fraud proofs during the dispute period. If this data is unavailable, fraudulent transactions cannot be challenged, jeopardizing the rollup’s security. ZK Rollups, while leveraging cryptographic validity proofs to ensure correct state transitions, still require transaction data to be accessible for auditing purposes and for users to verify the correctness of computations. Without robust DA mechanisms, rollups risk liveness failures — where new blocks cannot be produced due to missing data — or even finalizing invalid transactions. This makes DA essential for maintaining trustlessness and ensuring rollup operations remain secure and efficient.
Publishing transaction data directly on Ethereum ensures maximum security and decentralization but comes with prohibitive costs. Storing calldata on Ethereum’s mainnet is expensive due to high gas fees, making it economically unsustainable for high-throughput rollups. For instance, storing 1MB of calldata on Ethereum can tens of thousands of dollars, more than twenty-five hundred times more than it would have cost on rollups (Updegrave), limiting scalability for rollups that rely heavily on on-chain DA. To reduce costs, some rollups utilize external Data Availability Layers (e.g., Celestia or Avail) that store transaction data off-chain. While these layers improve scalability and reduce congestion on Ethereum, they introduce trust assumptions: a 51% or 67% attack on an external DA layer could render transaction data unavailable or inaccessible. Furthermore, Ethereum’s smart contracts verifying DA claims cannot independently confirm whether the off-chain data is genuinely available. This reliance creates vulnerabilities where malicious actors could collude with nodes in the DA network to withhold critical transaction data while submitting only block headers for finalization on Ethereum.
Liveness failures occur when rollup nodes lack access to the necessary transaction data to reconstruct the L2 state or produce new blocks. Even if transactions are settled on Ethereum, missing data can disrupt rollup operations and undermine user confidence in the system. This issue is particularly acute for Optimistic Rollups since validators cannot generate fraud proofs without access to complete transaction data. In addition, in Optimistic Rollups, validators require access to all transaction data during the dispute period to construct fraud proofs against invalid batches. If even a small portion of this data becomes unavailable during the dispute window, fraudulent transactions may go unchallenged, leading to incorrect state transitions being finalized.
The lack of standardized DA solutions across rollups creates fragmentation within the ecosystem. Each rollup may adopt its own DA strategy — such as publishing calldata on Ethereum or relying on bespoke DA layers — which hinders interoperability between different L2 networks and limits composability within the broader Ethereum ecosystem.
Approaches to Addressing DA Challenges
Publishing transaction data directly on Ethereum provides strong guarantees of trustlessness and decentralization by making all necessary information publicly available for verification. However, this approach incurs significant costs due to high gas fees and limits scalability for high-throughput applications. Modular blockchains like Celestia or Avail provide specialized DA solutions that store transaction data off-chain while maintaining accessibility through cryptographic attestations. These layers significantly reduce costs but introduce risks related to trust assumptions and potential attacks on the external DA network.
Some rollups adopt hybrid approaches by publishing critical metadata on Ethereum while storing bulk transaction data off-chain in external DA layers. This balances cost efficiency with security but requires careful design to mitigate liveness failures caused by partial data unavailability. Data availability sampling allows light nodes to verify block availability by sampling only a fraction of the block’s data rather than downloading it entirely. This method reduces resource requirements while ensuring that withheld data can be detected with high confidence. DAS has emerged as a promising solution for addressing scalability issues in modular DA layers.
Emerging rollup architectures are being designed with modular DA components that can be swapped out as needed. This allows rollups to choose between different DA solutions based on their specific requirements for cost efficiency, scalability, and security.
Do We Actually Care about (De)Centralization?
While addressing security and centralization issues in the rollup space is widely regarded as essential for ensuring the long-term viability of L2 solutions, there are arguments against prioritizing these concerns at the current stage of rollup development. These arguments suggest that the trade-offs associated with enhanced decentralization and security may not align with the immediate needs of users, developers, and the broader Ethereum ecosystem.
One argument is that rollups already inherit a significant degree of security and decentralization from L1. Since all transaction data and proofs are posted to Ethereum, users can exit rollups directly through L1 even if the rollup itself becomes compromised or censored. This reliance on Ethereum ensures that rollups benefit from its robust security guarantees, making them inherently more secure than traditional custodial or centralized solutions. Consequently, some argue that further decentralizing components such as sequencers may not be immediately necessary, as Ethereum’s settlement layer already provides a strong foundation for trustlessness and dispute resolution.
An analysis of the rollup ecosystem reveals that only one of the twenty largest rollups has reached Stage 2 maturity, with three at Stage 1 and the remaining sixteen at Stage 0. This distribution indicates that users prioritize transaction speed and cost efficiency over decentralization and advanced security mechanisms. Stage 0 rollups rely heavily on centralized components, such as sequencers, to ensure high throughput and low latency, aligning with user demands for quick and seamless transactions. However, this focus on performance comes at the expense of decentralization, a core principle of blockchain technology.
The dominance of Stage 0 rollups highlights the challenge of balancing usability with decentralization. While more mature rollups offer enhanced security and trustlessness, their higher costs and technical complexities can deter adoption. This trend underscores the need for developers to advance decentralization without compromising user experience. Failing to address these concerns risks undermining Ethereum’s foundational principles and its vision as a secure, decentralized platform for future applications.
Another perspective is that efforts to decentralize components such as sequencers or governance mechanisms can introduce significant complexity and performance trade-offs. Centralized sequencers currently enable higher throughput, lower costs, and faster transaction processing, which are critical for driving user adoption and ecosystem growth. For instance, centralized sequencers allow for efficient batching and ordering of transactions, facilitating scalability while maintaining a seamless user experience. In contrast, decentralization often leads to increased latency and reduced performance. Critics contend that prioritizing decentralization prematurely could hinder adoption by alienating users who value speed and cost efficiency over theoretical improvements in censorship resistance or fault tolerance.
Furthermore, many rollup projects adopt a phased approach to decentralization, focusing initially on building robust systems with centralized components before transitioning to more decentralized architectures. This strategy of “progressive decentralization” allows projects to iterate rapidly, attract users, and refine their technology without being burdened by the complexities of fully decentralized systems from the outset. For example, most rollups plan to eventually implement decentralized sequencers and governance mechanisms but prioritize addressing immediate user needs — such as scalability and low fees — overachieving full decentralization in the short term.
Practical challenges also arise in implementing full decentralization at this stage. Decentralized sequencers require novel infrastructure designs that balance trustlessness with functionality. Shared sequencing networks or decentralized models face limitations such as increased costs, coordination challenges among participants, and reliance on L1 throughput for cross-rollup interoperability. These hurdles make it difficult for rollups to achieve meaningful decentralization without compromising usability or economic sustainability.
Finally, it is argued that centralization in certain aspects of rollup design may be acceptable for specific use cases. Not all applications require maximum censorship resistance or fault tolerance. For example, gaming or social applications may prioritize performance over strict adherence to blockchain principles such as decentralization. In these scenarios, centralized components like sequencers may suffice to meet user requirements while maintaining low operational costs.
Conclusion
According to Vitalik Buterin, decentralization is needed for: “avoiding points of trust, minimizing censorship vulnerabilities, and minimizing centralized infrastructure dependency.” The road toward decentralization in the rollup space is of paramount importance and cannot be overstated. Rollups have emerged as a critical component of Ethereum’s scalability roadmap, allowing for off-chain transaction processing while maintaining the security of the main chain. However, as these technologies evolve, the next few years will likely set the foundation for Ethereum’s future. If developers fail to prioritize decentralization during this pivotal period, it risks undermining Ethereum’s core principles and its vision as a decentralized platform.
The importance of decentralization extends beyond security; it is also critical for fostering user participation and community-driven governance. Decentralized rollups allow users to actively contribute to network maintenance by running validator nodes or participating in governance mechanisms. This inclusivity not only strengthens the network but also aligns with Ethereum’s broader vision of democratizing access to financial and technological infrastructure. Projects like Taiko have demonstrated a clear commitment to decentralization through mechanisms such as contestable rollups and decentralized governance frameworks. These approaches ensure that rollups remain transparent and responsive to community needs while reducing reliance on centralized entities.
Despite these benefits, achieving full decentralization in rollups remains a complex challenge. Trade-offs between performance and security are inevitable; more decentralized systems may experience reduced throughput or higher costs due to increased resource requirements. However, these trade-offs can be mitigated through innovative solutions such as shared sequencing layers or hybrid models that balance efficiency with decentralization. For example, shared sorting layers allow multiple rollups to collaborate on transaction ordering while maintaining composability and interoperability across Layer 2 networks. Such solutions align with Ethereum’s ethos while addressing practical challenges associated with scaling.
The transition toward decentralized rollups is not merely a technical endeavor; it is a philosophical commitment to preserving the integrity of blockchain technology. Ethereum’s success as a platform for dApps and financial systems depends on its ability to uphold these principles. As highlighted by Vitalik Buterin’s proposed milestones for rollup decentralization, this process requires deliberate effort and long-term planning. Developers must prioritize building robust frameworks that incorporate decentralized governance, permissionless validation mechanisms, and transparent operations.
Written by Jason Yu (Research Team Fall 2024)
References:
Aranovich, Juan. “Should Ethereum Layer 2s Urgently Decentralize Their Sequencers?” Unchained. 6 Jun. 2024, https://unchainedcrypto.com/should-ethereum-layer-2s-urgently-decentralize-their-sequencers/
Awosika, Emmanuel. “Data Availability Or: How Rollups Learned To Stop Worrying And Love Ethereum.” Substack. 20 Dec. 2023, https://ethereum2077.substack.com/p/data-availability-in-ethereum-rollups
GrowThePie. https://www.growthepie.xyz/
Jun. “Making Sense of Based Rollups” Bankless. 26 Sep. 2024, https://www.bankless.com/based-rollups-primer
L2BEAT. Accessed 5 Dec. 2024, https://l2beat.com/scaling/summary
Mirpuri, Jash. “A Deep Dive into Shared Sequencers: The Future of Decentralization in the Blockchain Ecosystem.” Flagship. 12 Jul. 2023, https://flagship.fyi/outposts/market-insights/a-deep-dive-into-shared-sequencers-sequencer-networks/
Ngo, Tom. “L2 centralization is a ticking time bomb for blockchain.” Blockworks. 20 Aug. 2024, https://blockworks.co/news/layer-2-centralization-poses-dangers-for-blockchain
Updegrave, Henry. “Data Availability Comparison: Cost Per MB Posted for Ethereum Blobs Versus Celestia SuperBlobs.” Conduit. 23 Oct. 2024, https://blog.conduit.xyz/data-availability-costs-ethereum-blobs-celestia/