# SEACAST Delivers Ethereum’s 1 MiB Payload to 99% in 0.64 Seconds

Published 2026-10-21 · Crypto.world Team · Crypto.world

> Ethereum Research author cskiraly presents SEACAST, a broadcast design that delivered a simulated 1 MiB payload to 99% of 500 nodes in 0.64 seconds. The post compares its results with gossipsub and ECBroadcast and details limits in the simulations and formal model.

SEACAST delivered a simulated 1 MiB payload to 99% of 500 Ethereum nodes in 0.64 seconds, compared with 4.5 seconds for today’s gossipsub. In the same all-home network test, each node received an average of 1.3 MiB with SEACAST, against 4.2 MiB with gossipsub. [Ethereum Research](https://ethresear.ch/t/seacast-fast-lean-broadcast-for-ethereum-with-an-executable-spec-and-a-formal-model/26167) author cskiraly published the design on October 21, 2026, alongside an executable specification and a formal model.

The figures come from simulation, not a live Ethereum deployment. The comparison used 16 connections per node and ten simulated networks; the all-home setup assigned each node a 50 Mbps uplink and a 100 Mbps downlink. The post says SEACAST also beat ECBroadcast’s Reed–Solomon mode in those tests, completing the 99% threshold in 0.64 seconds versus 0.97 seconds while nodes received 1.3 MiB versus 3.7 MiB.

## Chunks move before the full payload arrives

SEACAST stands for Segmented, Encoded, Availability-sampled broadcast. Rather than make each node wait for an entire object before forwarding it, the protocol divides objects into coded chunks that peers can pass along as they arrive. Receivers can request missing chunks from one peer at a time, while senders announce what they hold and limit outstanding transfers according to their uplink.

The design also lets nodes receive only the data they need. A node may want an entire execution payload, its assigned data-availability columns, or selected cells for sampling; its forwarding duty is treated separately from that interest. That distinction aims to avoid sending every byte to every participant or creating a separate gossip topic for each piece.

“No broadcast can deliver such an object faster than its source’s uplink lets it out, or with fewer than one copy per node,” cskiraly writes.

> “No broadcast can deliver such an object faster than its source’s uplink lets it out, or with fewer than one copy per node.”
>
> [cskiraly, author of the Ethereum Research post](https://ethresear.ch/t/seacast-fast-lean-broadcast-for-ethereum-with-an-executable-spec-and-a-formal-model/26167)

## Sampling expands the design beyond payloads

The post describes SEACAST as a framework for several kinds of Ethereum data, including execution payloads, blob data-availability columns and rows, and potential future execution proofs or consensus blocks carrying post-quantum signatures. The researchers report a simulation of a column-shaped data object in which nodes received the eight columns assigned to their custody and forwarded no others. They explicitly caution that the test is a proxy, not PeerDAS: checking a cell was treated as free in the simulator.

Under a separate bandwidth test, the measured design carried a 1 MiB payload across 500 home nodes as uplinks fell from 50 Mbps to 10 Mbps. Completion time for 99% of nodes rose from 0.63 seconds to 1.8 seconds, while the delay stayed between 2.2 and 2.5 times the researchers’ calculated lower bound. Data traffic remained around 1.6 payload copies per node, with control traffic adding about a tenth of a copy.

## The specification runs, but the model found gaps

SEACAST’s specification writes protocol rules as Python-like code, and a small simulator executes those code blocks directly. The post distinguishes that executable document from the separate simulator used to produce the performance measurements; the latter uses a research implementation based on Prysm and a modified gossipsub library. An earlier version of the text was checked rule by rule against that implementation, and the comparison exposed two implementation errors that were later fixed.

The formal model, written in Quint and checked in part with the TLA+ model checker, examined small networks of three or four nodes and objects with at most two data and two parity chunks. It found ten liveness gaps involving delivery, promises and expiry. Version 0.5 fixes eight; two remain open, involving a node switching between blocks and a new connection not being offered data lost with an old connection.

## Live-network performance is still unmeasured

The post says the frequency of those gaps in real networks has not been measured, and its performance implementation predates the fixes. The formal checks also have defined limits: exhaustive checking at two data chunks has not completed, and the model omits heartbeat gossip, simultaneous objects and restarts. The next test is therefore not another simulated comparison, but whether the revised protocol and its remaining delivery rules hold up in deployed client networks.

## Sources

1. Original source [Ethereum Research](https://ethresear.ch/t/seacast-fast-lean-broadcast-for-ethereum-with-an-executable-spec-and-a-formal-model/26167) ethresear.ch

## Related stories

- [Ethereum Delays Glamsterdam’s Hoodi Upgrade Over Contract Risks](https://crypto.world/news/ethereum-delays-glamsterdam-hoodi-upgrade)
- [Ethereum Draft Sets Out Post-Quantum Keys for Validators](https://crypto.world/news/ethereum-draft-post-quantum-validator-key-registry)
- [Ethereum Draft Targets a Fourfold Cut in Finality Time](https://crypto.world/news/ethereum-eip-8453-decoupled-consensus-finality)
- [Ethereum Staking Hits 43.8 Million ETH as Yield Falls to 2.63%](https://crypto.world/news/ethereum-staking-43-8-million-eth-q3-2026)
- [Fhenix Takes CoFHE to Ethereum and Arbitrum Mainnet](https://crypto.world/news/fhenix-cofhe-ethereum-arbitrum-mainnet)

## Explore

[More articles](https://crypto.world/news)

---

Canonical: https://crypto.world/news/seacast-ethereum-broadcast-design-simulation
