You can find all published reports here:

Sunnyside Devnet Reports (Internal)

Overview

• 60 blobs/block under 22M gas usage (blobs+regular tx) on 120-node Interop Devnet. Additional gas usage and larger block size had minimal impact on the bandwidth usage and max blob throughput.
• Only 20 blobs/block with bandwidth limit under 22M gas usage (100/50Mbps Supernode & 50/25Mbps Fullnode as per EIP-7870)
• Perfect PeerDAS (16 8-validator nodes uniquely custodying each column) achieved 60 blobs/block under 22M gas usage & 30 blobs/block with bandwidth limits
• Genesis sync failed on Perfect PeerDAS network from Fulu epoch
• Found an issue with Teku’s backfill, being fixed by Teku. Waiting other CLs to support backfill.

Introduction

This week, we performed 2 major devnet tests under fusaka-devnet-2 spec.

Our previous tests only had blob transactions. We re-ran the previous tests with regular transactions load, consuming up to 50% gas. This will reveal how regular transactions affect the block propagation’s competition with the data column propagation across the nodes.

We also ran a Fusaka Devnet with only 16 nodes, each custodying 8 data columns, so that we can see how the network performs when each subnet gets custodied by exactly one client.

Genesis syncs were tested on both devnets with https://github.com/testinprod-io/kurtosis-sync-test.

We also tested validator custody backfill for Teku. Currently, only Teku is listed as supporting validator custody backfill - if your client is ready for backfill, we’ll be happy to test it in our devnet.

Devnet 1: Devnet-Interop

Setup & Configurations

Test Scenarios