A data availability (DA) layer is a specialized blockchain whose sole job is storing and serving the raw transaction data that rollups need for security verification. Instead of posting all data to expensive Ethereum L1 calldata, rollups can post to a cheaper dedicated DA layer (Celestia, EigenDA, Avail) that is optimized specifically for high-throughput data storage and sampling.
What Is a Data Availability Layer?
3 min read
The short version
Ethereum L1 is a premium vault: extremely secure but expensive to store things in. A DA layer is a purpose-built warehouse next door: cheaper to rent space, still secure enough for what you need, and anyone can walk in and check that your data is really there. Rollups store their receipts in the warehouse instead of the vault, saving money while keeping the ability to prove everything if needed.
How It Works
Why DA layers exist: Ethereum L1 charges rollups for every byte of data posted. Even with EIP-4844 blobs (cheaper than calldata), L1 data space is limited and expensive at scale. DA layers solve this by: (1) Providing much higher throughput (Celestia targets 8-32 MB/block vs. Ethereum ~130 KB of blob space per block). (2) Using data availability sampling (DAS) so nodes only download small random pieces to probabilistically verify all data exists. (3) Charging 10-100x less per byte than Ethereum L1. Major DA layers: Celestia (live, independent chain with its own validators, first dedicated DA layer), EigenDA (secured by restaked ETH via EigenLayer, inherits Ethereum economic security), Avail (Polygon-originated DA chain with its own validator set). Tradeoffs: using an external DA layer means trusting that chain's validator set for data availability (weaker guarantee than Ethereum L1). If the DA layer validators all go offline or collude to withhold data, rollup users might not be able to prove their state. The security assumption is: Ethereum L1 DA = maximum security (900K+ validators). External DA layer = good security (100-500 validators, smaller but still substantial economic stake).
Cost comparison: Ethereum blobs vs. Celestia for a rollup
A rollup processes 500 transactions per batch, each batch approximately 50 KB of compressed data. Posting to Ethereum blobs: blob fee market charges approximately $0.01-$0.10 per blob (128 KB max). Cost per batch: ~$0.05 average. Monthly (at 1 batch/minute): ~$2,160. Posting to Celestia: data costs approximately $0.001-$0.01 per 50 KB. Cost per batch: ~$0.005. Monthly: ~$216. Savings: 90% cheaper on Celestia. The rollup passes these savings to users as lower transaction fees. The tradeoff: if Celestia goes offline, the rollup cannot prove its state to Ethereum until Celestia recovers. For a gaming rollup with small transaction values, this tradeoff is acceptable. For a $10B DeFi protocol, they might prefer paying 10x more for Ethereum-native DA.
What People Get Wrong
DA layers replace Ethereum
DA layers complement Ethereum. Rollups still settle on Ethereum (post proofs and resolve disputes there). The DA layer only handles data storage. Settlement security still comes from Ethereum L1. The combination is: Ethereum for trust, DA layer for cheap storage.
Celestia is the only DA layer
Celestia was first to market but EigenDA, Avail, and NEAR DA all offer data availability services. Each has different security models, pricing, and throughput. The DA layer market is competitive and growing.
Using a DA layer makes a rollup less of a "real" rollup
The Ethereum community distinguishes: a rollup posting to Ethereum DA is a "rollup." One posting to an external DA is technically a "validium" (ZK) or "optimium" (optimistic). Both are valid scaling approaches with different security/cost tradeoffs. Neither is fake.
Keep Reading
Sources & Further Reading
- Celestia Documentation
Official docs for the first dedicated data availability layer
- EigenDA Overview
DA service secured by restaked ETH through EigenLayer
- L2Beat DA Comparison
L2Beat labels each rollup with its DA method for easy comparison
Questions People Also Ask
- How does data availability sampling work?
- Instead of every node downloading all data (which limits throughput), nodes download small random chunks. If enough random samples are available, there is a mathematical guarantee (via erasure coding) that the full data can be reconstructed. This lets DA layers support much larger data volumes without requiring every participant to have massive bandwidth.
- What happens if a DA layer goes offline?
- The rollup cannot post new data, which means new state updates cannot be verified by L1. Existing funds on L1 are safe (the settlement layer still works). Users might not be able to withdraw until the DA layer recovers and pending proofs can reference the data. For time-sensitive DeFi, this is a real risk. For casual usage, a few hours of DA downtime is tolerable.
- Should I care which DA layer my L2 uses?
- For most users: the practical difference is fees (external DA = cheaper transactions). For large deposits or DeFi usage: Ethereum-native DA provides the strongest guarantee that your data will always be available for withdrawal proofs. Check L2Beat for which DA method your L2 uses and its risk assessment.