Data Availability Sampling
How light nodes can be cryptographically confident that a block's full data was published — without downloading it — and why this is the key unlock for Ethereum's full danksharding and the modular blockchain stack.
The basics
Why Data Availability Matters
Rollups inherit security from their host chain by posting all their transaction data to that chain. The host chain just stores it; anyone can later download it to reconstruct the rollup's state. But what if the rollup operator publishes a state root claiming a certain outcome — and withholds the data needed to verify it? Without the data, no one can detect fraud. This is the 'data availability problem.' The host chain must guarantee that posted data is actually available to anyone who wants to check.
The Trivial Solution and Why It Doesn't Scale
The trivial solution is: every full node downloads every block, including all rollup data. Done. But this caps how much data the chain can carry — it's bounded by what an ordinary node can download. To 10x or 100x the data capacity, you'd need to make full nodes much more powerful, hurting decentralisation. So a more clever approach is needed.
Data Availability Sampling (the Idea)
Data availability sampling lets light nodes sample tiny random pieces of each block. If the data is fully available, every sample succeeds. If even a small fraction is withheld, with high probability some sample will hit a missing piece and the node will detect it. With erasure coding (a technique borrowed from storage systems), missing a small amount of data means the whole thing is unreconstructable — so a withholder either has to publish almost everything or nothing.
- Problem: how do we know a block's data was actually published?
- Trivial solution: every node downloads everything — limits capacity
- DAS: light nodes sample random pieces and use math (erasure coding) to be sure
- Result: more total data can be posted without making every node bigger
Key Takeaways
- Rollups need the host chain to guarantee posted data is available, not just stored
- Full downloads cap capacity; DAS unlocks scaling without sacrificing decentralisation
- Light nodes can verify availability with high confidence by sampling small random pieces
- DAS is the key technology behind Ethereum's full danksharding and the modular blockchain stack
Going deeper
The Erasure Coding Trick
DAS depends on a math trick: encode the original data into a longer code such that any large enough subset of the code reconstructs the original. Reed-Solomon codes are the classic example. If you encode 1 MB of data into 2 MB using Reed-Solomon, any 1 MB you recover lets you reconstruct the original. This means a withholder can't selectively hide 'just a few percent' — they have to hide more than half, which is loudly detectable. The math turns 'is the data available' into 'is at least half the code available' — a binary question that sampling can answer with high confidence.
What Light Nodes Actually Do
A DAS light node opens connections to many full nodes, requests random samples (small chunks of the erasure-coded block data), and checks each sample against a Merkle or KZG commitment. If all samples come back successfully, the node concludes — with statistical confidence depending on how many samples were taken — that the full data is available. The number of samples needed for, say, 99.99% confidence is small (a few dozen), regardless of block size. This scales: a node doesn't need bigger storage or bandwidth to verify a larger block; it just keeps the same sample count.
Celestia, EigenDA, Avail — The DA Ecosystem
Several modular data-availability layers have launched: Celestia (DAS-first, launched 2023), EigenDA (built on EigenLayer restaking), Avail (built on the Substrate framework), and others. They offer rollups a cheaper alternative to posting data on Ethereum. The trade-off: rollups posting on Ethereum get Ethereum's security; rollups posting on Celestia get Celestia's security. Some rollups use 'sovereign' setups where a single DA layer is the trust anchor; others use Ethereum and treat alternative DA layers as caches.
Ethereum's DAS Roadmap
Ethereum's proto-danksharding (EIP-4844, shipped 2024) did not implement DAS — every full node downloaded every blob. PeerDAS (EIP-7594) went live with the Fusaka upgrade on 3 December 2025, so nodes can verify blob availability by sampling rather than downloading entire blobs, and blob limits have since been raised through Blob Parameter Only forks (target 14, maximum 21 per block as of November 2026). Full danksharding, which adds further DAS so blob capacity can expand to ~32 MB/block, remains on the roadmap.
Key Takeaways
- Erasure coding turns 'is the data available?' into 'is at least half the code available?' — a binary question sampling can answer
- A light node samples a few dozen random chunks regardless of total block size
- Modular DA layers (Celestia, EigenDA, Avail) offer cheaper alternatives to Ethereum DA
- Ethereum shipped PeerDAS, a first form of DAS, in the Fusaka upgrade (3 December 2025); full danksharding remains a future upgrade and proto-danksharding was the precursor
Under the hood
Security Assumptions and Attacker Models
DAS provides probabilistic availability guarantees against a network with at least one honest sampling node. If every sample-serving node colludes to withhold the same data, sampling can't detect it. So DAS requires a sufficient, diverse set of sampling participants. This is why DAS-secured chains care so much about light-node count and geographic distribution — the security argument breaks down if the sampling network is too small or too centralised.
KZG Commitments + Polynomial Codes
Modern DAS designs use polynomial codes (a generalisation of Reed-Solomon) and KZG commitments. The block's data is interpreted as evaluations of a polynomial; the commitment is a constant-size cryptographic summary of the polynomial; each sample is a single evaluation point with a tiny opening proof. This makes the commitment + proof sizes very small even for large blocks, and is what makes DAS practical in real networks. The KZG trusted setup ceremony (~140,000 participants in 2023) was the foundation for both EIP-4844 and future DAS.
Fraud Proofs vs. Validity Proofs in a DAS World
DAS interacts with rollup proof systems. Optimistic rollups need a window during which fraud proofs can be submitted; if the data isn't available, fraud proofs can't be constructed, and the optimistic-rollup security model breaks. ZK rollups generate validity proofs from the data — if the data isn't available, the prover can't produce a proof and the rollup can't progress, but security isn't compromised. This is one structural argument for ZK rollups: they handle DA failures gracefully.
Beyond Rollups — Other DAS Use Cases
DAS isn't only for rollups. Any system where a publisher commits to large data and needs to prove availability can use it: decentralised file storage with availability guarantees, ML model weights with cryptographic auditability, large data feeds where availability matters. As modular DA layers mature, expect non-rollup uses to grow.
Watch the Number of Sampling Light Nodes
The security of a DAS-based system depends on how many sampling nodes are active. When evaluating a chain that depends on DAS (such as Celestia), look at the number of light clients participating — not just the validator count. Ethereum's PeerDAS is already live but works differently: each full node holds only a subset (1/8th) of the blob data. A small light-client set is a security concern that wouldn't show up in traditional metrics.
Key Takeaways
- DAS security depends on a sufficient and diverse sampling network — light-node count matters
- KZG + polynomial codes make commitments and opening proofs small, making DAS practical
- DAS interacts with rollup proof systems: ZK handles DA failures more gracefully than optimistic
- Beyond rollups: DAS unlocks new patterns for any data-publication system needing availability proofs
- Treat 'light-client count' as a key security metric for DAS-based chains
Related Content
Related Coins
Related Topics
Key Terms
2026 Trends
More Topics
Browse all 189 topicsBitcoin ETFs & Institutional Crypto
How spot Bitcoin ETFs work, what institutional adoption means for the market, and why owning ETF shares is fundamentally different from owning the coin.
UTXO vs Account Model
How Bitcoin's UTXO model differs from Ethereum's account model — what each gets right, what each makes harder, and why the choice shapes everything from parallelism to replay protection.
EIP-4844 and Proto-Danksharding
How Ethereum's March 2024 Dencun upgrade introduced blob space — a new transaction type that cut rollup data costs by ~10x and set the stage for full danksharding.
Account Abstraction (ERC-4337)
How ERC-4337 brings smart-contract wallets, social recovery, gasless transactions, and arbitrary signature schemes to Ethereum without changing the protocol — and why every wallet UX of the next decade will use it.
Finality and Fork Handling
What it means for a transaction to be 'final,' why different chains finalise differently, how reorgs happen, and the MEV incentives that complicate the picture.
State Pruning and Expiry
Why Ethereum's node storage needs have grown to terabytes, what 'archive,' 'full,' and 'light' nodes actually store, and how proposals like EIP-4444 and new state-tree designs (verkle or binary) aim to keep node hardware accessible for years to come.
ECDSA vs EdDSA vs BLS
The three signature schemes you'll meet across the crypto stack — what each does well, what tradeoffs they impose, and why Ethereum uses three of them simultaneously.
zk-SNARK vs zk-STARK vs PLONK
How the three major proof-system families compare on trusted setup, proof size, prover cost, and quantum resistance — and which production rollups picked which.
References & further reading
- primary
- secondary
- primaryEIP-7594: PeerDAS - Peer Data Availability Sampling
Specification of Ethereum's live data availability sampling (status: Final).
- primaryethereum.org - Fusaka
Confirms Fusaka and PeerDAS went live on 3 December 2025.