Data availability sampling (DAS) is a technique that lets resource-limited clients gain high statistical confidence that all the data behind a block has been published, without downloading the whole block. Each client requests a few random fragments of an erasure-coded dataset; if enough randomly chosen fragments are returned across many clients, the full data can be reconstructed, so withholding it becomes detectable. DAS is a foundational primitive for scalable, modular blockchains and rollup-centric architectures.
- Data availability sampling lets a light client gain near-certainty that a block’s data has been fully published without downloading all of it, by requesting random fragments of an erasure-coded dataset. It implements Data Availability guarantees and is a key enabler of Scalability for Rollup and Sharding designs.
Overview
- For a blockchain to be safely verified, the data behind each block must be available so that anyone can check its validity. A block producer who publishes a header but withholds the data could hide invalid state.
- Downloading every block in full does not scale to high throughput. Data availability sampling breaks this trade-off by allowing many small clients to collectively verify availability.
- The block data is first erasure-coded, so the original can be reconstructed from any sufficiently large subset of fragments. This makes withholding all-or-nothing: hiding even a little data requires hiding a large fraction.
- Each client samples a few random fragments. If a producer withholds data, most clients will, with high probability, request at least one missing fragment and detect the fault.
Mechanisms
- Erasure coding, often Reed-Solomon over a two-dimensional layout, expands the data with redundancy so any large-enough subset suffices for reconstruction.
- Polynomial commitments such as KZG bind the encoding so that fragments can be verified against the block header without trusting the producer.
- Random sampling across many independent clients drives the probability of undetected withholding exponentially toward zero.
- Networked retrieval and reconstruction allow honest participants to recover the full data once enough fragments circulate.
Applications
- Light-client verification of large blocks on modular and sharded chains.
- Dedicated data availability layers such as Celestia that sell verifiable availability to rollups.
- Scaling Ethereum through danksharding-style blob data verified by sampling.
- Underpinning Rollup security where execution is off-chain but data must remain available.
Key aspects
- DAS gives probabilistic, not absolute, guarantees, but the confidence approaches certainty with modest sampling.
- It depends on a healthy peer-to-peer network to serve and reconstruct fragments.
- It pairs with fraud or validity proofs to deliver full verification of off-chain execution.