Ethereum rollups process transactions away from Mainnet but still need a place to publish enough data for independent verification. As blob capacity grows, requiring every node to download every blob would eventually make bandwidth a barrier to running a node. PeerDAS changes that model: nodes sample and custody different pieces of blob data instead of every node carrying everything.
Quick answer: PeerDAS, introduced on Mainnet with the Fusaka upgrade in December 2026, uses data-availability sampling, erasure coding, and a distributed gossip network. It gives Ethereum strong probabilistic evidence that rollup data is available while reducing the amount an ordinary node must download.
Why rollups need data availability
A rollup executes transactions on a layer 2 network and posts commitments and transaction data to Ethereum. The data allows users and independent software to reconstruct the rollup state, verify transitions, and, depending on the design, challenge invalid behavior or produce proofs.
Data availability does not mean that Ethereum executes every rollup transaction. It means the data needed for verification has been published and can be retrieved. If an operator posted only a state root while withholding the underlying data, users might be unable to reconstruct balances or exit safely.
Our Ethereum guide explains the base network. PeerDAS specifically addresses the bandwidth required to make rollup data available.
From calldata to blobs
Early rollups published transaction data as calldata in ordinary Ethereum transactions. Calldata competes with execution activity for block space and is stored as part of chain history.
The Dencun upgrade introduced EIP-4844 blob transactions. Blobs are a specialized, temporary data format designed for rollups. Ethereum consensus verifies commitments to blob data, while the data itself does not become permanent EVM state. This made rollup data cheaper and gave it a separate fee market.
Before PeerDAS, each node still downloaded each blob. Raising blob counts therefore raised every node’s bandwidth roughly together. That “everyone downloads everything” rule protected availability but limited scaling.
What data-availability sampling changes
Data-availability sampling (DAS) divides the work among nodes. Each node requests a small, randomly assigned portion of the data and verifies it against a cryptographic commitment. If many independently distributed nodes successfully receive their samples, the network gains high confidence that the full dataset is available.
Sampling needs redundancy. Otherwise a block producer could publish only the exact pieces that some nodes request and hide the rest. PeerDAS uses Reed-Solomon-style erasure coding to extend blob data with additional recoverable information. The original can be reconstructed when enough of the extended data is present.
How PeerDAS works
1. Blobs are extended
Blob data is represented mathematically and evaluated at additional points. This doubles the evaluations and creates redundancy. KZG commitments let nodes verify that a received piece belongs to the committed blob.
2. Extended data is split into columns
Ethereum’s PeerDAS design divides extended blob data into 128 columns. These columns are distributed through dedicated peer-to-peer gossip subnets.
3. Ordinary nodes custody a subset
A regular node participates in at least eight randomly selected column subnets. Instead of receiving all 128 columns, it receives a subset determined from its node identity. Different nodes cover different subsets, distributing storage and bandwidth across the network.
4. Validators check availability before voting
Validators follow fork-choice rules that require successful data-availability checks before accepting and voting for a block. The result is part of consensus behavior, not merely an optional analytics service.
5. Supernodes provide full custody and repair
Nodes connected to validators with a combined balance at or above the protocol’s supernode threshold subscribe to all column subnets. Nodes that reconstruct missing data can redistribute recovered columns, helping heal gaps.
Why the capacity gain is probabilistic
An individual ordinary node does not see every piece. Availability confidence comes from random distribution, cryptographic verification, redundant encoding, many independent samplers, and consensus rules. The security calculation depends on participation and network assumptions; it is not the same as one node storing a full copy.
Ethereum.org describes an 8× theoretical blob-capacity limit relative to the previous download-everything model. “Theoretical” matters: safe production capacity is raised gradually and monitored rather than switched instantly to the maximum.
Blob-Parameter-Only forks
Fusaka introduced Blob-Parameter-Only (BPO) forks so Ethereum can adjust blob targets and limits without waiting for a full feature upgrade. The network began with conservative parameters, then scheduled stepwise increases while client teams observed bandwidth, propagation, and stability.
This means PeerDAS activation and maximum blob capacity are different milestones. Users should not assume that every theoretical increase became available on Fusaka day one. Current blob parameters should be checked in official network documentation.
What PeerDAS means for rollup users
- More room for data: higher blob capacity can reduce congestion in the blob fee market.
- Potentially lower L2 fees: data cost is one component of a rollup fee, so more supply can help—but execution, proving, sequencing, and application charges still matter.
- Home-node viability: sampling limits the bandwidth increase imposed on every ordinary node.
- Gradual effects: fee changes depend on demand and BPO parameter increases, not only on the existence of PeerDAS.
What PeerDAS does not solve
PeerDAS is not a universal solution to rollup risk. It does not remove:
- smart-contract bugs in rollup bridges;
- centralized sequencer downtime or censorship;
- upgrade-key and governance risk;
- faulty proof systems or challenge mechanisms;
- user errors when bridging assets; or
- differences between rollups, validiums, sidechains, and other scaling designs.
Data availability is one layer of the security model. A user evaluating an L2 should also inspect the stage of decentralization, exit mechanism, upgrade delay, sequencer, proof system, and bridge.
PeerDAS vs full Danksharding
PeerDAS is an intermediate step toward Ethereum’s longer-term Danksharding vision. It uses one-dimensional erasure coding per blob. Future designs may use broader two-dimensional coding and additional protocol changes. Those future features should be treated as research or roadmap items until activated on Mainnet.
Frequently asked questions
Did PeerDAS launch on Ethereum Mainnet?
Yes. The Ethereum Foundation’s 2026 protocol update states that Fusaka brought PeerDAS to Mainnet in December 2026.
Does every validator store all blob data?
No. Regular nodes sample and custody subsets of columns, while designated supernodes subscribe to all column subnets.
Does PeerDAS make L2 transactions free?
No. It expands data-availability capacity. Rollups still incur data, execution, proving, sequencing, and infrastructure costs, and user fees respond to demand.
Are blobs permanent?
Blob data is ephemeral under Ethereum’s protocol design. Its purpose is to be available long enough for rollup verification, not to serve as permanent general-purpose storage.
This article reflects official Ethereum materials reviewed on 31 July 2026. Protocol parameters can change through scheduled upgrades.