mailwal

[ethresear.ch] [Networking] SEACAST: fast, lean broadcast for Ethereum, with an executable spec and a formal model

Csaba Kiraly
cskiraly
October 29

Update, 29 October 2026: this post’s living version, and four new measurements. This post stays as published. Its text is now kept current on notes.ethereum.org, and each version is a tag of eth-networking-studies with the data and scripts behind it. Version 1 adds:

  • Figures 4 and 5 on the fixed code. The time stays within the spread between runs; the bytes fall a little on the thinner links.
  • FullDAS in Figure 4, with every node sampling every cell. On the all-home network it falls between SEACAST and ECBroadcast; SEACAST is faster and lighter than it on each of the ten networks of both kinds.
  • A second simulator. Under Shadow, SEACAST is as fast or faster, with about the same bytes, and stays faster and lighter than today’s gossipsub on every network. Given Shadow’s queue, our simulator ranks the protocols the same way.
  • Larger networks. At 1000 and 2000 nodes, SEACAST takes about a tenth longer than at 500, with about the same bytes.
Figure 4
Figure 4, updated: today’s gossipsub, ECBroadcast’s Reed–Solomon mode, FullDAS with every node sampling every cell, and SEACAST in one simulator: 500 nodes with 16 connections each (FullDAS also opens short-lived connections for its samples), a 1 MiB payload, ten simulated networks per panel. Each small point is one protocol on one network; the large markers are the medians over the networks; open circles on a dashed line are the earlier posts’ first three tiers. Time is plotted on a log scale.

SEACAST’s specification has moved on as well, to version 0.8; its changelog lists what each version changed.

Visit Topic to respond.

You are receiving this because you enabled mailing list mode.