PANDAS: Peer-to-peer, Adaptive Networking for Data Availability Sampling within Ethereum Consensus Timebounds
Matthieu Pigaglio, Onur Ascigil, Michał Król, Sergi Rene, Felix Lange, Kaleem Peeroo, Ramin Sadre, Vladimir Stankovic, Etienne Rivière · 2025
Read PANDAS: Peer-to-peer, Adaptive Networking for Data Availability Sampling within Ethereum Consensus Timebounds by Matthieu Pigaglio, Onur Ascigil, Michał Król, Sergi Rene, Felix Lange, Kaleem Peeroo, Ramin Sadre, Vladimir Stankovic, Etienne Rivière (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2507-00824.txt.
Introduction
Matthieu Pigaglio and eight co-authors present PANDAS for data availability sampling in Ethereum in two thousand twenty-five. Layer two data must spread and be sampled within four seconds. That bound comes from the consensus slot clock of the chain. No past system met it at planetary scale in their account.
PANDAS uses light, direct peer exchanges to spread and sample data. It needs no change to consensus or discovery protocols. Loss, node failure, and silent peers are all handled in design. A thousand node prototype and larger simulations test the build. Sampling met the deadline under global latency in those runs.
Catalog record
| CREATOR | Matthieu Pigaglio, Onur Ascigil, Michał Król, Sergi Rene, Felix Lange, Kaleem Peeroo, Ramin Sadre, Vladimir Stankovic, Etienne Rivière |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 78 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2507-00824
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
PANDAS: Peer-to-peer, Adaptive Networking for Data Availability Sampling within Ethereum Consensus Timebounds Matthieu Pigaglio, Onur Ascigil, Michał Król, Sergi Rene, Felix Lange, Kaleem Peeroo, Ramin Sadre, Vladimir Stankovic, Etienne Rivière Abstract . Layer-2 protocols can assist Ethereum’s limited throughput, but globally broadcasting layer-2 data limits their scalability. The Danksharding evolution of Ethereum aims to support the selective distribution of layer-2 data, whose availability in the network is verified using randomized data availability sampling (DAS). Integrating DAS into Ethereum’s consensus process is challenging, as pieces of layer-2 data must be disseminated and sampled within four seconds of the beginning of each consensus slot. No existing solution can support dissemination and sampling under such strict time bounds. We propose PANDAS, a practical approach to integrate DAS with Ethereum under Danksharding’s requirements without modifying its protocols for consensus and node discovery. PANDAS disseminates layer-2 data and samples its availability using lightweight, direct exchanges. Its design accounts for message loss, node failures, and unresponsive participants while anticipating the need to scale out the Ethereum network. Our evaluation of PANDAS’s prototype in a 1,000-node cluster and simulations for up to 20,000 peers shows that it allows layer-2 data dissemination and sampling under planetary-scale latencies within the 4-second deadline. 1. Introduction Ethereum, the largest blockchain supporting smart contracts, currently supports adding fewer than a few tens of transactions per second to its main (layer-1) chain. Complementarily to layer-1 scalability improvements ( hafid2020scaling ; rebello2024survey ) , expanding support for layer-2 protocols ( gangwal2023survey ) is now a priority for the Ethereum community ( ethereum2024roadmap ) . Layer-2 protocols such as side chains and rollups have the potential to process large amounts of transactions ( gangwal2023survey ; jourenko2019sok ) . These protocols periodically produce compressed or batched layer-2 transaction data, which they make available via the layer-1 blockchain. For instance, participants in an optimistic rollup can download this data, verify its correctness, and submit fraud proofs ( ethereum2024optimistic_rollups ; kalodner2018arbitrum ; optimism ) . The throughput of layer-2 protocols depends on how much data they can attach to layer-1 blocks. Previously, the only solution was adding layer-2 data as costly calldata transactions. These transactions competed for permanent block space with other layer-1 transactions, such as ETH transfers and DeFi interactions. In March 2024, the EIP-4844 ( Proto-Danksharding {}^{\text{1}} ) proposal ( ethereum_protodanksharding_eip4844 ) introduced the notion of blobspace . Layer-2 data can now be shared as opaque binary objects (blobs). Blobs are broadcast separately and referenced by blob-carrying transactions in the block, which include cryptographic commitments to their content. Nodes participating in consensus verify these commitments and make blob data available to layer-2 participants for a limited time (4,096 epochs, \sim 18 days). While improving over calldata transactions regarding costs and supported volume, EIP-4844 still requires blob data to be broadcast and received by all nodes. A significant upcoming step towards scaling support for layer-2 data in Ethereum is implementing data availability sampling (DAS). The Ethereum Danksharding 1 1 1 Danksharding is named after Dankrad Feist, an Ethereum researcher. Proto-danksharding is named after him and Diederik Loerakker ( protolambda ). ( ethereum2024roadmap_danksharding ) roadmap plans to support up to 32 MB blob data referenced by each layer-1 block. To avoid broadcasting this volume of data globally, blob data is erasure-coded, split, and distributed as collections of cells , so that each node holds only a fraction of the data. This shift introduces a new challenge: no single node can independently verify the availability of the full blob. To address this, Ethereum plans to adopt data availability sampling (DAS), wherein nodes collect random sample cells from the network until they reach overwhelming confidence that the complete data can be reconstructed. The Danksharding parameters imply sending 140 MB of erasure-coded cells to the network, with each node randomly sampling 73 cells (40 KB). Integrating DAS and Ethereum consensus is a challenge. The need for each node to collect randomly selected samples results in a multitude of exchanges over high-latency links. At the same time, Ethereum consensus imposes tight time constraints. A new block is generated every 12 seconds. A committee must validate each new block within the first four seconds after its creation. To avoid changes to the consensus protocol, DAS must also be completed within four seconds, allowing committee members to attest block validity and blob data availability simultaneously. Several approaches are under discussion for integrating DAS with Ethereum consensus, e.g., EIP-7594 (PeerDAS) ( eip-7594 ) , SubnetDAS ( subnetdas ) , or FullDAS ( fulldas ) . These leverage peer-to-peer networks already used by other Ethereum functions, particularly GossipSub ( gossipsub ) , a broadcasting network that pre-establishes dissemination overlays and uses them as channels for gossiping data. There is a discrepancy between the high costs of establishing new channels and the randomized nature of DAS. Multi-hop gossiping also has inherently high latency. These two factors lead these proposals to suggest that random sampling happens after the validation of blocks by the committee, i.e., past the 4-second deadline. Completing sampling after committee validation may reverse validation decisions due to the delayed detection of blob data unavailability. This requires modifications to Ethereum’s consensus to account for such decision-revert possibilities. This also impacts finality (i.e., how long before a block can be considered immutable) as pending data availability verifications delay decisions. Finally, the possibility of reverting past consensus opens the door to new attacks based on ex-ante reorganizations ( d2024recent ) . Contribution. We demonstrate that DAS can be integrated with the existing Ethereum consensus while meeting the Danksharding objectives. Dissemination and sampling of blob data can occur within the first four seconds of a consensus slot. This allows the committee to confirm blob data availability and block data correctness simultaneously and removes the need to adapt Ethereum’s consensus to delayed availability decisions. We present PANDAS, a peer-to-peer protocol that supports DAS in Ethereum. PANDAS builds upon the following key features: • It aligns with recent Ethereum evolutions, including Proposer-Builder Separation (PBS) ( ethereum2024roadmap_pbs ; heimbach2023ethereum ) , which introduces powerful builders responsible for preparing block and blob data and ordinary proposers elected by Proof-of-Stake consensus ( merge-ethereum ) . PANDAS leverages builders for efficient seeding of blob data. • It uses Ethereum nodes to host and sample blob data using peer-to-peer interactions. In contrast to other proposals ( eip-7594 ; subnetdas ; fulldas ) , PANDAS employs direct (one-hop) communication, using connectionless networking (UDP). Interactions adapt to nodes’ unavailability and faults, meeting the 4-second deadline in adverse environments or under inconsistent network views by different participants. • PANDAS supports Ethereum’s objectives of openness, decentralization, and scalability. Nodes’ and builders’ bandwidth requirements align with the typical capacities of home servers and cloud instances, and do not increase with system size. We implement PANDAS over libp2p ( libp2p ) , the network stack of the Ethereum Geth client ( geth ) , and deploy 1,000 nodes on an 80-server cluster using representative emulated WAN latencies. Additionally, we utilize a simulator whose results are cross-validated against prototype deployments. This enables us to confidently explore results for up to 20,000 nodes. Our evaluation shows that PANDAS meets the 4-second sampling deadline at all nodes at moderate scales and for the vast majority