Data availability is the property that required blockchain data has been published and is accessible to participants who need it. For a rollup, this can mean obtaining the data needed to reconstruct account state or independently examine proposed updates. A hash or other commitment identifies data, but does not provide the missing bytes or guarantee that someone will serve them.
How availability is checked
Nodes can download the required data directly. Data availability sampling instead checks selected portions of an encoded dataset. Erasure coding adds redundancy, allowing recovery from enough available pieces. Successful sampling provides probabilistic confidence under the scheme's parameters and network assumptions, rather than having every participant download everything.
Ethereum uses PeerDAS for blob data availability. Other networks, including Celestia, use their own sampling and publication designs.
Availability is distinct from correctness and archiving
Data can be available even when the transactions it describes are invalid. Conversely, a validity proof of an encoded computation does not make withheld data retrievable.
Ethereum blobs have a limited protocol retention period; separate archives can preserve copies afterward. Availability therefore concerns timely publication and retrieval, not a promise that every node stores every byte forever. Missing rollup data can obstruct state recovery, challenges or withdrawal proofs even when an onchain commitment remains visible.