Data Availability (DA) Solutions: Celestia, EigenDA, Avail & Danksharding (EIP-4844)

The Mathematics of Erasure Coding: 1D vs. 2D Reed-Solomon Matrix Codes

Historically, validating data availability required every node to download $100\mathrm{\%}$ of the block data. This created a severe scaling limit: block sizes could not exceed the bandwidth capacity of standard home validator internet connections (~1-2 MB per block).

To allow lightweight nodes (such as mobile devices and browser extensions) to verify data availability without downloading the entire block, modern DA architectures utilize Reed-Solomon Erasure Coding.

An $(n,k)$ Reed-Solomon code over a finite field takes $k$ original data chunks and expands them into $n$ encoded chunks ($n=2k$, a $2\times$ expansion factor) via polynomial interpolation. The fundamental mathematical property of Reed-Solomon coding: Any $k$ out of the $n$ encoded chunks are sufficient to reconstruct the entire original data set using Lagrange interpolation.

1. 1D Reed-Solomon Coding:

If a block contains $k$ data shares, the proposer encodes it into $2k$ shares. A malicious proposer attempting to hide even a single transaction must withhold more than $50\mathrm{\%}$ of the total shares ($k+1$ shares).

1. 2D Reed-Solomon Matrix Coding (Celestia & Danksharding):

To prevent a malicious block producer from generating an invalid erasure code (where shares are incorrectly computed), Celestia organizes the block data into a $k\times k$ matrix.

• Each row is independently erasure-coded into a $2k$ row using a separate Reed-Solomon polynomial.
• Each column of the resulting extended matrix is independently erasure-coded into a $2k\times 2k$ square matrix.
• The block header commits to the Merkle Roots of all $2k$ rows and $2k$ columns.

Because the matrix is bound in two dimensions, an attacker cannot generate an inconsistent code without corrupting an entire quadrant of the matrix. Furthermore, verifiers can prove an erasure coding fraud with an interactive or succinct ZK fraud proof of a single row or column.

Data Availability Sampling (DAS): Sub-Linear Random Query Verification

With 2D Reed-Solomon erasure coding in place, light nodes can verify data availability with overwhelming mathematical certainty using Data Availability Sampling (DAS).

The DAS algorithm executes as follows:

1. When a new block header is received, the light node does NOT download the full megabytes of block data.
2. The light node generates $S$ random sampling coordinates $(r_i,c_i)$ across the $2k\times 2k$ matrix.
3. The light node sends peer-to-peer requests to the network asking for the specific data chunk at $(r_i,c_i)$ along with its associated Merkle Inclusion Proof to the row/column root in the block header.
4. If a block producer is withholding data, they must withhold at least $25\mathrm{\%}$ of the entire 2D matrix (otherwise the network can reconstruct all withheld shares).

The probability $P_{\text{fail}}$ that a cheating block producer successfully hides data without the light node detecting a missing sample after $S$ independent random queries is:

$$P_{\text{fail}}\le {(1-\frac{1}{4})}^S={\left(\frac{3}{4}\right)}^S$$

For $S=20$ random samples (requiring only ~30 KB of total downloaded data):

$$P_{\text{fail}}\le {\left(\frac{3}{4}\right)}^{20}\approx 0.00317(99.68\mathrm{\%}\text{ confidence})$$

For $S=30$ samples:

$$P_{\text{fail}}\le {\left(\frac{3}{4}\right)}^{30}\approx 0.000178(99.982\mathrm{\%}\text{ confidence})$$

The profound architectural implication: as more light nodes join the peer-to-peer network and perform random sampling, the collective network can safely reconstruct and support massive 64 MB to 1 GB blocks without requiring any single node to download more than a few kilobytes. This decouples blockchain throughput from individual node hardware limits.

Celestia Architecture: Sovereign Rollups & Namespaced Merkle Trees (NMTs)

Celestia is the pioneer of dedicated modular consensus and data availability layers. Celestia purposefully strips away all execution capabilities: it contains no smart contract virtual machine, no state execution engine, and no application layer. It solely orders transactions and provides data availability guarantees.

Key architectural innovations in Celestia:

1. Namespaced Merkle Trees (NMTs):
2. In a traditional Merkle tree, data leaves are ordered arbitrarily.
3. In an NMT, leaves are partitioned and sorted by a specific 8-byte Namespace ID ($NID$).
4. Each internal tree node commits to the range of namespaces $[NI{D}_{\text{min}},NI{D}_{\text{max}}]$ contained in its subtrees.
5. This allows a specific rollup (e.g., Rollup A with $NID=0x42$) to query and download only the transactions belonging to its specific namespace, along with a logarithmic Merkle proof verifying that no Rollup A transactions were omitted from the block, saving $99\mathrm{\%}$ of download bandwidth.

1. Sovereign Rollups:
2. Unlike Ethereum rollups that settle on an on-chain smart contract, a Sovereign Rollup on Celestia uses Celestia purely as an ordered data pipe.
3. The canonical fork choice and fraud proof verification are executed locally by the rollup's own peer-to-peer node network (Client-Side Verification), granting the rollup full sovereign governance without requiring Layer-1 bridge hard-forks.

## EigenDA & Avail: Restaked Security vs. KZG Polynomial Commitments

Beyond Celestia and Ethereum Blobs, two major high-throughput DA protocols have emerged:

1. EigenDA (Restaked Hyperscale DA):
2. Developed by EigenLabs, EigenDA is built as an Actively Validated Service (AVS) on EigenLayer, leveraging over $15B+ in restaked Ethereum economic security.
3. Architecture: Decouples data availability from consensus. EigenDA does not run a sovereign PoS consensus chain. Instead, an off-chain Disperser splits data into erasure-coded chunks, generates KZG polynomial commitments, and distributes chunks to registered EigenLayer restaked node operators.
4. Performance: Eliminates consensus serialization bottlenecks, achieving up to 10 MB/s to 50 MB/s data throughput with fractional-cent fee costs.

1. Avail (Avail DA - Founded by Anurag Arjun, ex-Polygon):
2. Avail combines KZG Polynomial Commitments with 2D Reed-Solomon erasure coding.
3. Unlike Celestia (which uses fraud proofs to verify that erasure coding was computed honestly), Avail uses KZG commitments. Because KZG commitments are mathematically binding, light nodes can immediately verify erasure coding correctness without waiting for fraud proof challenge windows.
4. Avail integrates a unified interoperability messaging substrate (Avail Nexus) and a shared security network (Avail Fusion).

## Comparative Economic & Throughput Matrix across DA Layers

Selecting a Data Availability layer involves evaluating economic costs, settlement latency, security models, and light client trust assumptions.

While Ethereum Blobs remain the gold standard for high-value financial rollups requiring absolute settlement security, external DA layers have made micro-transaction consumer rollups economically viable.

Frequently asked questions

Why can light nodes not perform Data Availability Sampling on traditional blockchains like Bitcoin?

Traditional blockchains do not erasure-code blocks. Without 2D Reed-Solomon erasure coding, hiding a single 100-byte transaction requires withholding only 0.005% of the block, which light nodes cannot reliably detect without downloading almost the whole block.

What is a Namespaced Merkle Tree (NMT)?

An NMT is a Merkle tree where data is sorted by namespace ID. It allows rollups to download and verify only the data belonging to their specific chain without downloading transactions from other rollups sharing the same DA block.

How does EigenDA differ from Celestia?

Celestia is an independent sovereign Layer-1 Proof-of-Stake consensus chain. EigenDA is an Actively Validated Service (AVS) on Ethereum that uses restaked ETH for security and relies on KZG commitments without running a separate PoS consensus loop.

Related reading

• rollup data posting to external DA layers — Understand why Layer-2 rollups require secure data availability to prevent state freeze attacks.
• EigenLayer restaked validator security in EigenDA — Review the shared security architecture powering EigenDA restaked nodes.

## Sources

• EIP-4844: Shard Blob Transactions (Proto-Danksharding) — Ethereum Improvement Proposals
• Celestia documentation: data availability sampling — Celestia

No contributor to this article holds a professional cryptography or security credential. Every technical claim above is sourced to primary protocol documentation rather than to personal authority — follow the sources and verify anything you intend to act on.

Not financial advice. Crypto assets are volatile and can lose value. This article describes how protocols work, not what you should buy.