Data Availability Layers: Celestia and EigenDA in Layer 2 Scaling

Introduction

When a rollup posts a batch of transactions, the single most expensive thing it does is publish data. Not execution, not proving—publishing. For a Layer 2 to be trustless, anyone must be able to reconstruct its state, which means the underlying transaction data has to be available somewhere verifiable. On Ethereum, that "somewhere" was historically calldata, and it dominated rollup costs—often 80–90% of the total fee.

Data Availability (DA) layers exist to break this bottleneck. Instead of paying L1 execution-grade prices to store data, rollups publish to a purpose-built layer optimized for one job: guaranteeing that data was published and can be retrieved. Celestia and EigenDA are the two most prominent designs, taking very different architectural routes to the same goal.

This article covers the scaling problem DA layers solve, the technical architecture of both systems (data availability sampling vs. a restaking-secured network), their performance characteristics, ecosystem traction, the developer experience of integrating them, and the trade-offs you accept when you move data off Ethereum's base layer.

The Scaling Challenge

A monolithic Layer 1 like pre-rollup Ethereum bundles execution, consensus, settlement, and data availability into every node. Every full node re-executes every transaction and stores all data. This is maximally secure but caps throughput: Ethereum settles roughly 15–30 transactions per second, and gas competition pushes fees to dollars or tens of dollars during congestion.

Rollups address execution by moving computation off-chain and posting compressed batches plus validity or fraud proofs back to L1. But they still must publish their transaction data so the network can verify state transitions and so users retain the ability to exit without permission. If data is withheld, a malicious operator can freeze funds or produce unverifiable state—the classic data withholding attack.

This is the data availability dimension of the scalability trilemma. You can have low cost, high throughput, or strong security guarantees—optimizing all three simultaneously is the hard part. Posting full data to Ethereum L1 maximizes security but is expensive. Posting to a centralized server is cheap but reintroduces trust. DA layers stake out a middle ground: dedicated networks that provide cryptographic or cryptoeconomic guarantees of availability at a fraction of L1 cost.

EIP-4844 (proto-danksharding) gave Ethereum its own native answer—blobs—dropping rollup data costs dramatically. But blob capacity is finite (~target 3, max 6 blobs/block at ~128 KB each), and high-throughput chains still hit ceilings. That's the gap external DA layers fill.

Technical Architecture

Celestia: Data Availability Sampling

Celestia is a sovereign, Cosmos-SDK-based blockchain whose only purpose is ordering and guaranteeing the availability of data—it does not execute rollup transactions. Its core innovation is Data Availability Sampling (DAS) combined with 2D Reed-Solomon erasure coding.

When a block producer publishes a block, the data is arranged into a k × k matrix and erasure-coded into a 2k × 2k extended matrix. The property that matters: the original data can be reconstructed from any 75% of the extended shares. To hide data, a malicious producer would have to withhold more than 25% of the shares—and that becomes statistically detectable.

Light nodes exploit this. Rather than downloading the full block, each light node requests a handful of random shares plus their Merkle proofs against the block's data root:

For each light node:
  for i in 1..n:                       # n samples, e.g. 16
    (row, col) = random_coordinate()
    share, proof = request(row, col)
    assert verify(share, proof, data_root)
  # If all n samples succeed, availability
  # confidence ≈ 1 - 0.75^n  (>99% at n=16)

With enough light nodes sampling independently, the network collectively verifies that the entire block is available without any single node downloading it. Namespaced Merkle Trees (NMTs) let each rollup fetch only its own data by namespace, so a rollup ignores other chains' blobs.

Rollups integrate via a Blobstream bridge (formerly Quantum Gravity Bridge), a smart contract on Ethereum that relays Celestia data roots. The rollup's settlement contract checks that its batch data was committed to Celestia before accepting a state update. Security model: Celestia's own Tendermint/CometBFT proof-of-stake consensus plus the DAS guarantee. Trust assumption is that Celestia's validator set is honest and that enough light nodes sample. It is a separate trust domain from Ethereum.

EigenDA: Restaking-Secured DA

EigenDA takes a fundamentally different approach. It is an Actively Validated Service (AVS) built on EigenLayer, inheriting security from Ethereum validators who restake their ETH. There is no separate consensus chain for ordering; EigenDA is a network of operators that store data and produce signed attestations.

The flow:

1. A rollup's disperser erasure-codes a blob and splits it into chunks.
2. Chunks are distributed to operators, each storing only a fraction.
3. Operators return KZG polynomial commitment signatures attesting they hold their chunk.
4. Signatures are aggregated (BLS) into a single proof posted to an EigenDA manager contract on Ethereum.

// Simplified attestation verification on L1
function confirmBatch(
    BatchHeader calldata header,
    BlobVerificationProof calldata proof
) external {
    require(verifyAggregatedSig(header.sig, header.operators));
    require(stakedAmount(header.operators) >= quorumThreshold);
    confirmedBatches[header.batchRoot] = block.number;
}

The guarantee is cryptoeconomic: operators who fail to serve data can be slashed, with their restaked ETH at stake. Unlike Celestia, EigenDA (in its initial form) does not perform DAS on-chain; data availability rests on the assumption that a sufficiently staked quorum of honest operators retains the data and the disperser behaves correctly. KZG commitments let anyone verify a retrieved chunk matches what was committed, but the availability guarantee is economic, not sampling-based.

Both designs share the goal—cheap, verifiable data—but Celestia leans on cryptographic sampling and its own consensus, while EigenDA leans on Ethereum-aligned restaking economics.

Performance Metrics

DA throughput is measured in bandwidth (MB/s), not raw TPS, because the layer doesn't execute—the effective TPS depends on the rollups consuming it.

Metric	Celestia	EigenDA	Ethereum blobs (4844)
Throughput (current)	~8 MB/block (~1.3 MB/s effective)	~10–15 MB/s (targeting higher)	~0.375 MB/s target
Block/confirm time	~6 s blocks	~ Ethereum-anchored confirmation	~12 s slots
Cost per MB	Cents to low dollars	Comparable, often lower	Variable, higher under load

Celestia launched with ~2 MB blocks and has raised capacity over time (8 MB and rising), with a roadmap toward 1 GB blocks. EigenDA's published targets reach into tens of MB/s as operator count and bandwidth scale.

For a rollup, the practical impact is dramatic: data costs that were the dominant fee component fall by one to two orders of magnitude versus L1 calldata. An L2 batch that cost dollars in calldata can cost fractions of a cent in DA fees, translating into sub-cent end-user transactions. Finality nuance matters: with external DA, a rollup's "soft" confirmation follows the DA layer's block time, while "hard" finality still depends on the rollup's settlement on Ethereum (and, for optimistic rollups, the challenge window).

Ecosystem & Adoption

Celestia (mainnet "Lemniscate," live since late 2023) anchors a growing set of rollup frameworks and chains. Rollup-as-a-Service providers—Caldera, Conduit, AltLayer—offer Celestia DA as a one-click option, and stacks like the OP Stack (via Plasma/alt-DA mode), Arbitrum Orbit, and Polygon CDK support it. Notable consumers include modular chains and gaming/app-specific rollups where cheap data dominates economics. Celestia's TIA token secures the network and is used to pay blob fees.

EigenDA, launched on mainnet in 2024, benefits from EigenLayer's large restaked-ETH base (multiple billions in TVL at peak) and operator set. Early integrators include rollups built with the OP Stack and Arbitrum-derived stacks seeking Ethereum-aligned security without separate-token assumptions. Its adoption narrative leans on "stay within the Ethereum trust zone."

Developer activity for both is concentrated in the RaaS layer—most teams don't integrate DA from scratch; they pick it from a deployment menu. This means adoption tracks the RaaS providers' defaults and incentive programs as much as raw technical merit. TVL figures for DA layers are less meaningful than for L2s directly, since value sits on the rollups, not the DA layer; the more telling metrics are blobs posted per day and number of live chains consuming each layer.

Developer Experience

Neither layer requires you to abandon the EVM. Both plug in beneath an existing rollup stack, so your contracts and tooling (Hardhat, Foundry, ethers/viem) are unchanged. What changes is the rollup node's DA backend.

For Celestia, a node runs against a Celestia light or bridge node and submits blobs via the API:

celestia blob submit \
  0x42690c204d39600fddd3 \    # namespace
  "0x<batch-data>" \
  --node.store ~/.celestia-light

Many OP Stack / Orbit deployments expose this as a config flag rather than direct API calls—you set the DA provider and namespace, and the batcher handles submission. EigenDA integration similarly happens through a DA client/proxy sidecar that the rollup's batcher talks to over gRPC, dispersing blobs and returning a certificate the rollup posts to L1.

Documentation quality is solid on both sides, with Celestia's "build a sovereign rollup" tutorials (Rollkit) being especially developer-friendly, and EigenDA providing integration guides tied to specific rollup stacks. Migration from an L1-DA rollup is mostly a configuration and re-deployment exercise plus a security review of the new bridge/attestation contracts—not an application rewrite. The main new operational concern is running or relying on a DA node and monitoring its liveness.

Trade-offs & Limitations

The central trade-off is trust domain. Posting data to Celestia or EigenDA means your rollup's safety now depends on a system other than Ethereum. With Celestia, that's a separate validator set and its own token's economic security; a deep reorg or validator failure there affects your chain. With EigenDA, you stay Ethereum-aligned via restaking, but you inherit EigenLayer's still-maturing slashing mechanics and the assumption that the disperser and operator quorum behave honestly—EigenDA's launch design lacked on-chain DAS, leaning on cryptoeconomics instead of cryptographic sampling.

Centralization concerns apply to both: Celestia's validator set, like any young PoS chain, is more concentrated than Ethereum's; EigenDA's disperser has been a centralized component in early phases. Withdrawal/exit times are governed by the rollup, not the DA layer, but using external DA can complicate trustless exits—if the DA layer censors or stalls, users may be unable to reconstruct state to force-exit, which is why "alt-DA" modes are sometimes labeled as having weaker guarantees than full Ethereum DA.

Finally, fragmentation: multiple DA layers plus Ethereum blobs split liquidity, tooling, and security budgets across ecosystems, and bridges between trust domains are themselves attack surface.

Competitive Landscape

The DA market is a three-way contest. Ethereum blobs (EIP-4844, then full danksharding) are the security gold standard—same trust zone, no extra assumptions—but currently the most capacity-constrained and, under load, the most expensive. Celestia competes on raw cost and scalable throughput via DAS, attractive to app-specific and high-volume chains willing to accept a separate trust domain. EigenDA competes on Ethereum-alignment, pitching restaking security as a middle path: cheaper than blobs, more Ethereum-native than Celestia.

Other entrants—Avail (also DAS-based, Polygon-origin) and Near DA—crowd the field further. Celestia's unique advantage is being the first production DAS layer with a mature modular narrative; EigenDA's is leveraging the largest pool of restaked ETH. Positioning largely comes down to a builder's answer to one question: how much do you value staying inside Ethereum's trust boundary versus minimizing cost?

Roadmap & Future

Celestia's roadmap centers on scaling block size toward 1 GB blocks (the "Gigablock" vision) through larger blocks and more light nodes, plus improvements like lazy bridging and interoperability between rollups sharing its DA. EigenDA is expanding throughput targets, decentralizing its disperser, and tightening EigenLayer's slashing and security guarantees as the broader restaking ecosystem matures. The potential game-changer for both is full danksharding on Ethereum—if L1 DA capacity scales enough and stays cheap, the economic case for external DA narrows, pushing Celestia and EigenDA to compete harder on cost, throughput, and developer ergonomics.

Conclusion

Data availability layers solve the dominant cost in rollup economics by separating "is the data published?" from "does the chain execute it?" Celestia delivers cryptographic availability guarantees through data availability sampling and its own consensus, offering high throughput at low cost but introducing a separate trust domain. EigenDA delivers cryptoeconomic guarantees via restaked ETH, staying Ethereum-aligned at the cost of relying on operator honesty and maturing slashing. Choose Celestia when cost and scalable throughput dominate and a separate trust assumption is acceptable—typical of app-chains and high-volume rollups. Choose EigenDA when Ethereum-alignment is a hard requirement. And reach for Ethereum blobs directly when maximal security outweighs cost. The right answer is workload-specific, and the landscape is still moving fast.

---

Disclaimer: This article was written with AI assistance and edited by the author. It is for informational purposes only and does not constitute financial, investment, or trading advice. Always conduct your own research and consult with qualified professionals before making any investment decisions. Cryptocurrency investments carry significant risk and may result in loss of capital.

Published via NeuralKalym - Automated crypto content system