Data Availability Sampling (DAS)

Data Availability Sampling (DAS) is a technique that lets a node confirm, with very high statistical confidence, that all of a block's data has been published, by downloading only a few small random pieces instead of the whole block.

DAS relies on erasure coding. Before publishing, block data is extended with redundant information so that the full data can be rebuilt from only a portion of the pieces. A block producer trying to hide even a small part of the data would therefore have to withhold a large share of the pieces. Light nodes then request many random pieces from the network. If every request is answered, it becomes extremely unlikely that a meaningful amount of data is missing.

This matters because blockchains, rollups and sharded networks need their data to be openly available so anyone can check transactions or rebuild the chain's state. DAS lets ordinary devices take part in that check without downloading everything. Networks can then carry more data, such as rollup blobs, without forcing every participant to run expensive hardware. Celestia is built around DAS, and Ethereum's scaling plans for blob data also rely on it.

For example, suppose a producer withholds half of the extended data. Each random sample then has a 50% chance of failing. If a light node takes 30 samples and all succeed, the chance of being fooled is roughly one in a billion.

A common misconception is that DAS checks whether transactions are valid. It does not. It only shows that the data was published. Validity still depends on fraud proofs or validity proofs. DAS also assumes enough honest nodes are sampling, and that a node is not cut off from honest peers.

Learn more

Related terms