Antler by Autoflux
HomeArbitrum NitroData Availability
Report an error

Data Availability

Rollup vs. AnyTrust data modes, the Data Availability Committee, and how batch data reaches L1.

Path: data-availability

Third-party documentation. This is independently authored analysis of the public Arbitrum Nitro codebase — not the official docs, and not reviewed or endorsed by the Arbitrum Nitro team.

Data Availability

Like every rollup, Nitro must guarantee that the data needed to reconstruct L2 state is available to anyone who wants to verify it. Nitro's distinctive feature is that it ships in two data modes from one codebase: full rollup (data on Ethereum) and AnyTrust (data entrusted to a committee).

Rollup mode (Arbitrum One)

  • The sequencer compresses ordered batches and posts them to SequencerInbox on L1.
  • Originally as calldata; since the blob upgrade, as EIP-4844 blobs — the same tradeoff as other rollups: cheaper per byte, but data is only guaranteed present for ~18 days at the node level.
  • Data availability is therefore Ethereum's guarantee: anyone can rebuild the chain from L1 transactions at any time during the retention window.

AnyTrust mode (Arbitrum Nova)

AnyTrust relaxes the "data on Ethereum" requirement with a Data Availability Committee (DAC):

  1. The sequencer sends batches to the DAC, which signs a certificate attesting it has stored the data (members are known, such as major infra providers).
  2. The certificate — a short signature proof, not the data itself — is posted to L1's SequencerInbox. Data availability now rests on the DAC having kept the data, with a fraud-proof fallback for liveness.
  3. Security depends on a quorum of the DAC being honest. Nitro caps the damage a malicious majority can do (they can withhold data, triggering a liveness issue, but they cannot forge state or steal funds).

This drops DA costs by roughly an order of magnitude — Nova exists because high-frequency, low-value applications (e.g. gaming) don't need full rollup-grade DA.

The trust model difference

  • Rollup (One): data availability = Ethereum. No committee.
  • AnyTrust (Nova): data availability = honest-DAC assumption (data exists) + Ethereum (state settlement and fraud proofs).

How data is packaged

Both modes share the same packaging: L2 blocks are compressed into batches, framed for L1 inclusion, and tracked in a Merkle accumulator in SequencerInbox. The only difference is whether the frame bytes land on Ethereum (rollup) or in the DAC's store with a certificate on Ethereum (AnyTrust).

Edge cases

  • If the DAC withholds data, blocks referenced by an unserved certificate can't be reconstructed — but anyone who had the data can still verify, and validators will refuse to settle an assertion built on missing data.
  • Data fees differ dramatically between modes — an application on Nova pays far less per byte of calldata. Choose the chain whose DA guarantee matches your application's value profile.
  • After L1 blob expiry, nodes rely on archival providers or their own storage; a full node must either retain blobs or have a strategy (snapshots, archive nodes) for historical re-derivation.