Overview

We ran a series of tests to identify the root cause of peerDAS’s blob limitation.

We identified blob limits for the following client pairs:

• lighthouse/prysm with geth: 20 blobs/block

• lighthouse/prysm with All EL: 17 blobs/block

• lighthouse/prysm/lodestar + geth: 12 blobs/block

• lighthouse/prysm/teku + geth: 25 blobs/block

• lighthouse/prysm/grandine + geth: 9 blobs/block

• lighthouse/prysm/nimbus + geth: N/A, see below

Of the above results, our finding is:

• currently, some execution clients have txpool limits that cannot handle high blob counts & have good getBlobsV1 performance
• more analysis is needed into why grandine and lodestar has lower blobs/block throughput compared to other clients
  • run only single consensus client devnets to remove every external factors

The testing methodologies are identical to the previous tests.

Lighthouse / Prysm + Geth

• Nodes & Images
• Grafana