Intro to EigenDA: Hyperscale Data Availability for Rollups

If you're interested in integrating your rollup with EigenDA, please fill out the EigenDA questionnaire!

EigenDA is a secure, high throughput, and decentralized data availability (DA) service built on top of Ethereum using the EigenLayer restaking primitive. Developed by EigenLabs, EigenDA will be the first actively validated service (AVS) to launch on EigenLayer. Once launched, restakers will be able to delegate stake to node operators performing validation tasks for EigenDA in exchange for service payments, and rollups will be able to post data to EigenDA in order to access lower transaction costs, higher transaction throughput, and secure composability across the EigenLayer ecosystem, with security and throughput designed to horizontally scale with the amount of restake and operators opted into servicing the protocol.

We intend for EigenDA to contribute the following to the Ethereum ecosystem:

• An innovative DA solution for rollups that builds toward the Ethereum scaling endgame, with security derived from, and value contributed back to, Ethereum stakers and validators. EigenDA is built on some of the core ideas as well as libraries underlying Danksharding, which is a crucial upgrade to the Ethereum ecosystem, and can play a role in battle-testing these technologies.
• A standard for high throughput and low cost that enables the growth of new on-chain use cases. EigenDA will enable applications across multiplayer gaming, social networks, and video streaming, with a flexible cost model for both variable and fixed fees.
• A key ingredient for protecting decentralization. In shared security systems like EigenLayer, if every node operator is required to download and store every chain that uses the system, very few node operators will be able to keep up and the system has the risk of eventually centralizing. EigenDA is built to prevent this centralizing tendency; it achieves high-performance while distributing work across many participating nodes, requiring each operator to do only a small amount of work.
• A proof point for the power of programmable trust. EigenDA seeks to demonstrate that Ethereum stakers and validators can support critical Ethereum infrastructure in addition to Ethereum consensus, and that AVSs (like EigenDA) and AVS users (like rollups using EigenDA) can succeed with new business and token models based in modularity on top of the Ethereum trust network.

We are excited that several teams have specified plans to integrate EigenDA into their L2 infrastructure, including: Celo as they transition from L1 to Ethereum L2; Mantle and its suite of complementary products out of the BitDAO ecosystem; Fluent providing a zkWASM execution layer; Offshore providing a Move execution layer; Layer N providing zk-OP hybrid rollups designed for financial applications; and many more.

We continue to look for foundational partners to build on EigenDA, and invite rollups of all kinds to reach out and learn more about the protocol - especially those that are striving to enable high throughput use cases that require 10 MBps throughput and higher. If you’re interested, please get in touch via this form: EigenDA Questionnaire.

Technical Architecture

The diagram below shows the basic flow of data through EigenDA.

1. The rollup Sequencer creates a block with transactions, and sends a request to disperse the data blob.
2. The Disperser is responsible for erasure encoding data blobs into chunks, generating a KZG commitment and KZG multi-reveal proofs, and sending the commitment, chunks, and proofs to the operator nodes of the EigenDA network.
3. Rollups will be able to run their own disperser, or use a dispersal service that a third party (for example, EigenLabs) operates for convenience and amortization of signature verification costs. It is possible for a rollup to use a dispersal service optimistically, such that in the case the service is non-responsive or censoring, the rollup can use its own disperser as a backstop, thus getting amortization benefits in the optimistic mode without sacrificing censorship resistance.
4. The EigenDA nodes verify the chunks they receive against the KZG commitment using the multireveal proofs, persist the data, then generate and return a signature back to the Disperser for aggregation.

Technical Considerations

Now that we have a basic understanding of EigenDA’s architecture, let’s discuss the benefits and properties that the system is designed to achieve. Below is a short list of properties we believe are necessary for a good and useful data availability layer for rollups:

• Economics
• Throughput
• Security
• Customizability

We’ll address each property from the EigenDA perspective.

Economics

Today, many L2s use Ethereum as their data availability layer due to its crypto-economic security guarantees. This leads to extremely high and volatile costs, as rollups compete with all other Ethereum users for limited blockspace based on congestion pricing. For example, both Arbitrum and Optimism have spent tens of millions of dollars on calldata (data availability) costs on Ethereum year-to-date, with no consistency month-to-month. One of the primary value propositions of DA systems is to dramatically lower these costs and give rollups greater predictability in their cost structure.

Understanding Cost Reduction

There are three fundamental dimensions of cost incurred in operating a DA system. Let’s analyze how EigenDA aims to minimize the underlying cost structure on each dimension:

• The capital cost of staking. To stake capital in order to secure a DA layer, stakers may want to receive a certain percentage yield in order to offset their opportunity cost. EigenDA reduces the capital cost of staking by using EigenLayer, which employs a shared security model that allows the same stake to be utilized across a variety of applications, creating an economy of scale.
• Operational cost. Instead of requiring each node to download and store all data, EigenDA uses erasure coding to split data into smaller chunks, and requires operators to download and store only a single chunk, which is a fraction of the full data blob size. This imposes a lower cost on each operator as compared to storing the full blob, making EigenDA “lightweight” to operate by many nodes. As more nodes join the EigenDA network, the resource costs incurred by every node on the network decreases. This enables EigenDA to be secured by a large set of operators at low and marginally decreasing cost, enabling a philosophy of abundance rather than scarcity.
• Congestion cost. As bandwidth utilization in any blockchain approaches the system’s capacity, data starts to get more expensive in order to combat congestion. EigenDA reduces congestion in two ways: 1, by having a higher throughput, it seeks to make congestion a rare phenomenon; 2, by allowing for bandwidth reservation, EigenDA can guarantee rollups pre-reserved throughput at a discounted cost. To maintain flexibility, EigenDA also allows rollups to pay for throughput on-demand.

Rollup Economics

Rollup economics differ fundamentally from those of an L1 because DA costs are not only high and unpredictable, but they are incurred in a non-native token. This makes it difficult for rollups to make price commitments to users and subsidize initial adoption, because they must take “exchange rate risk” between their own rollup token and the token in which DA fees are paid. In comparison, an L1 pays a fixed amount of inflation and can offer a certain amount of transactions-per-second for free to attract users.

In EigenDA we are exploring mechanisms for rollups to pay stakers in native rollup tokens for a predictable long-term reservation rate, on terms that EigenLayer restakers find acceptable. This couples the inherent scale advantages of a shared security system with the inherent advantage of stable native token payments, in order to help bootstrap usage of rollups.

Throughput

Throughput is another foundational value proposition of DA systems. EigenDA is designed to achieve horizontal scaling such that the more operators there are on the network, the more throughput the network enables. In private testing with a set of 100 nodes with standard performance characteristics, EigenDA has demonstrated a throughput of up to 10 MBps, and has a roadmap to scale to 1 GBps. This opens the door for bandwidth-intensive applications like multiplayer gaming and video streaming on top of Ethereum.

EigenDA delivers high throughput via three pillars in its design:

• Decoupling DA and consensus. Existing DA systems couple the attestation of availability of data blobs with the ordering of data blobs into a “monolithic” architecture. Data attestation is parallelizable as nodes can independently attest to the availability of distinct data blobs; however, ordering requires serialization of the data blobs, thus creating significant consensus lags. While this coupling can be beneficial to security in systems which are designed to be the final source of ordering, it is neither required nor useful in a DA system designed to accompany the Ethereum blockchain, which has its own ordering system that rollups rely on for settlement anyway. By shaving off the unnecessary complexity of ordering and designing a pure DA system, EigenDA improves significantly on throughput and latency.
• Erasure coding. EigenDA enables rollups to take the data they want to post to EigenDA, decompose it into smaller chunks, and perform erasure coding on those chunks before storing the data as fragments. Using KZG polynomial commitments (a mathematical scheme at the heart of ZK proofs), EigenDA requires nodes to only download small amounts of data [O(1/n)] rather than download entire blobs. Unlike systems that use fraud proofs to detect malicious incorrect coding of data, EigenDA employs validity proofs in the form of KZG commitments, which enables nodes to verify correct coding of the data.
• Direct Unicast instead of P2P. Existing DA solutions use a peer-to-peer (P2P) network for transmitting blobs, in which operators receive blobs from their peers, then re-broadcast the same blob to others. This significantly constrains the achievable DA rate. In EigenDA, a Disperser sends blobs directly to EigenDA’s operators. By relying on unicast (direct communication) for dispersing data, EigenDA can confirm DA at native network latency rather than via an expensive gossip protocol. This eliminates the significant gossiping penalties that come with P2P, and results in faster data commitment times.

Security Features

We use security as an umbrella term that covers safety and liveness, as well as decentralization and censorship resistance. The following features attest to EigenDA’s security:

• EigenLayer. By using restaking, EigenDA borrows two distinct aspects of security from the EigenLayer system: 1, economic security; 2, decentralization. EigenDA is built to take advantage of these two distinct elements of trust from the EigenLayer and Ethereum ecosystem in a synergistic manner.
• Proof of Custody. A key mode of operator failure in EigenDA is that nodes sign off on data items without actually storing them for the required period of time. To solve this problem EigenDA uses a mechanism called proof-of-custody, which was originally proposed by Justin Drake and Dankrad Feist of the Ethereum Foundation. With proof-of-custody each operator must routinely compute and commit to the value of a function which can only be computed if they have stored all the blob chunks allocated to them over a designated storage period. If they attest to blobs without computing this function, the ETH held by the node can be slashed by anyone who has access to their data item.
• Dual Quorum. EigenDA also has a feature called Dual Quorum, where two separate quorums can be required to attest to the availability of data. As an example, one quorum would be composed of ETH restakers (the ETH quorum), and the second quorum could be composed of stakers of the rollup’s native token. Both quorums are treated as an independent and redundant source of DA, so both quorums have to be compromised before EigenDA can fail.
• Censorship Resistance. EigenDA offers higher instantaneous censorship resistance than coupled DA layers. This is because coupled DA architectures usually rely on a single leader or block proposer to linearly order the data blobs, thus creating an instantaneous censorship chokepoint. In contrast, in EigenDA, rollup nodes can directly disperse and receive signatures from a majority of EigenDA nodes, thus improving the censorship resistance to a majority of EigenDA nodes rather than being constricted by a single leader.

Security Analysis

As discussed, EigenDA is built on ETH staking via EigenLayer and uses erasure codes with a configurable coding ratio that can be set by rollups. There are three distinct perspectives to analyze the security of a blockchain system like EigenDA; we describe each perspective and how it applies to EigenDA as described above:

Byzantine fault tolerance (BFT): Assume some fraction of nodes are honest and follow the protocol exactly, and some are malicious and can deviate from the protocol arbitrarily.

• EigenDA is safe, i.e. data can be retrieved, as long as X% of nodes are honest, where X can be 10% to 50% depending on the coding rate.

Nash Equilibrium model: Analyzes the economic incentives of each node or small collusion of nodes to follow the protocol, assuming that nodes across distinct collusions act independently.

• As long as the collusion size is smaller than (1-X), storing and serving data to users is the Nash Equilibrium: storing data is ensured as an equilibrium by proof-of-custody, which slashes the ETH of nodes that do not store the data; serving data is ensured as an equilibrium because data is dispersed across many nodes, inducing a competitive market to serve data.

Pure cryptoeconomic model: Assumes all stake is held by the same node and models the cost of economic corruption.

• As long as the data is available, or equivalently, as long as X% of nodes are honest, the ETH staked by any node that doesn’t custody data will be slashed. However, EigenDA does not have unconditional cryptoeconomic safety; if all nodes collude and withhold data, then it may not be possible to slash them. In the Dual Quorum model described earlier, where both ETH and a native rollup token are staked, the rollup can slash the native token even in cases where it is not possible to slash ETH.

As we can see, EigenDA is built on a trust model that requires not only economic trust derived from ETH staking, but also decentralization and independence of operators in order to function safely. Fortunately, EigenLayer allows EigenDA to borrow both of these trust mechanisms from Ethereum.

Customizability

Rollup developers can implement EigenDA with flexibility to adjust parameters as desired. The modular nature of EigenDA enables rollups to customize safety/liveness tradeoffs, staking token modalities, erasure coding, payment tokens accepted, and more.

As discussed in earlier sections some of the most important flexible decisions in EigenDA are economic. For example, a rollup can choose to use dual quorum staking, in which their own token is staked to guarantee data availability; or a rollup can choose a cost structure that is either on-demand or pre-reserved.

Strategic Considerations

Finally, (and thanks for sticking with us) we believe EigenDA offers strategic value to rollups outside of the technical properties described above.

• Ethereum stakers and validators are the beating heart that powers EigenLayer, and accordingly, that powers EigenDA. By adopting EigenDA, rollups may align with these Ethereum stakeholders, who expressly value decentralization, censorship-resistance, open access software, and composable, permissionless innovation.
• EigenDA is planned to be the first of many AVSs to launch in the EigenLayer ecosystem. We foresee that as the number of AVSs grows, there will be composability benefits between them, which benefit the end-users of those AVSs, which we expect to include rollups of many kinds. For example, after EigenDA, we expect to see AVSs launch with use cases including sequencing, fast confirmation, watcher networks, bridging, fair ordering, and even artificial intelligence.
• EigenDA is in the early stages of a long journey towards achieving a big vision. The EigenLabs team is seeking foundational partners who want to work closely with us on EigenDA, but also for the long haul, across many projects and as long-term collaborators. We hope this represents an opportunity to drive the Ethereum ecosystem towards more open innovation together, and we look forward to supporting your project as much as we possibly can.

The Path Forward

Consistent with our design philosophy we have a phased roadmap for EigenDA, where various features of EigenDA will be built and shipped incrementally. The first version of EigenDA will go to testnet later this year.

If you’re interested in participating, please get in touch with our team via this form: EigenDA Questionnaire. We look forward to meeting you!