PeerDAS Breakout Call #23 | Apr 1, 2025

Protocol Calls & happenings
1

Apr 1 (Tues), 2025, 14:00 UTC
Duration: 60 minutes
Zoom link
Ethereum Protocol Call calendar
Notes
Previous call
Homepage: DAS.wtf

Agenda Overview

A. Client Updates
B. Devnet / Testing Updates
C. Spec / EIP Discussion
D. Open Discussion

Agenda Details

Client Updates

PRIORITY: All teams to focus on implementing getBlobsV2 and running across clients

CL Client Teams

Team Link Call 22 Updates Timely Discussion Topics
Prysm link • Worked on redesigning the database to fix issues with data column storage
• Added missing unit tests for PeerDAS
• Worked w/ Francis from Base worked on getBlobV2 and equivalent API when proposing a block		 • Provide an update on the status of bug
Lighthouse link • Created spec PR for distributed publishing
• Working on making Lighthouse compliant with the spec		 • Update on sync issues
• Update on: ethpandaops/lighthouse:peerdas-devnet-6-089338d
Teku link • Released validator custody, started testing with DevOps
• Added new PeerDAS metrics with help from Katya
• Working to resolve various bugs		 -
Lodestar link • Made progress with refactoring and unit testing
• Onboarding a new contributor to help with validator custody
• Addressing peer management issues to ensure there are peers listening to all subnets
• Looking at improving how peers connect to those listening to correct topics		 -
Nimbus - • Working to improve getBlobsV1 performance (empty & partial responses from EL)
• Worked with Sunnyside to accommodate up to 24 blobs
• Finished independent column syncer, starting testing ASAP
• Improved peer pool and peer mgmt		 • FOCUS FOR WEEK: finish validator custodry, start on getBlobsV2
Grandine - - • Completed cell_proofs integration based on spec with the compute_cells methods just implemented by the team
• Refactored on reconstruction and block acceptance flow to reconstruct after block accepted
• Added withhold-data-columns-publishing flag

EL Client Teams

Team Link Call 22 Updates Timely Discussion Topics
Geth - • Open pull request about encoding, limited progress due to vacation schedules • Update on bloobpool implementation status
Nethermind - • Has proof of concept branch with API methods implemented, ready to join DevNet as execution layer client • Execution Engine Error: Nethermind encountered array length errors when processing engine_getPayloadV5 requests during block production with Lighthouse

• Root Cause Identified: Error stemmed from incompatible blob transaction proof formats - client encountered proof-type-v1 transactions (Deneb format) when expecting proof-type-v2 (PeerDAS format with 128 proofs per blob)

• Implemented Solution: Modified Nethermind to ignore older proof-format transactions when building Osaka/Fulu blocks (may result in temporarily empty blocks during testing)

• Specification Update: Confirmed that new proof types should use wrapper_version 0x01 instead of 0x02 as previously implemented
Reth - • All types implemented, waiting for specs to be locked in, 1-2 days from being ready -
Besu - - -
Erigon - - -

Devnet / Testing Updates

peerdas-devnet-6 (specs)

Devnet-6 Open Question:

  • Devnet-6: Who is going to work on EELS/EEST for proof computation to tx sender?

STEEL Updates

  1. EL changes are limited to the network wrapper that encapsulate blob transaction information and blob data into a gossipable object between peers. This can be tested via Hive’s Engine simulator. Tests will be added / updated here.
  2. As there are no changes to the EVM or EL block header, noo changes are required to EELS execution-specs or EEST’s consensus test set. Therefore:
  • No peerdas-devnet-6 specific release will be made from EEST
  • EL clients can proceed with testing using the latest develop release (provided their branches are based on Prague) without waiting for a specific peerdas-devnet-6 release from EEST.

Sunnyside Updates

Sunnyside Devnet Updates - 04/01 (link)
Findings from this week:

  • Different consensus clients show varying blob throughput capacities with Lighthouse handling up to 40 blobs/block, Teku 30 blobs/block, while Prysm and Lodestar manage 25 blobs/block, and Grandine 20 blobs/block.
  • The main bottleneck appears to be declining getBlobs hit rates from execution clients, causing increased column reconstruction work and network instability.
  • Most consensus clients are performing adequately without relying on getBlobsV2.
    Testing strategy for next week:
  • Investigate implementation differences causing throughput variances between clients
  • Continue debugging Geth’s blobpool behavior
  • Test performance with distributed blob publishing
  • Conduct tests without EL blob gossiping to measure getBlobs impact
  • Run comparative tests with other execution clients using unlimited blobpool sizes.
  • Work on standardizing metrics across clients to enable better cross-client performance comparisons.

Spec / EIP Discussions

P0 : Open Spec PRs

  • EIP | Update EIP-7594: include cell proofs in network wrapper of blob txs (one last comment)
  • consensus-specs | Transition period from 4844-style blob transactions to PeerDAS-style blob transactions (link)
  • cell proof computation | Add EIP-7594 (PeerDAS) related changes (beacon-api) (builder-specs) (engine-api)
  • execution-spec-tests | feat(tests): peerdas tracking issue (guessing this is not a blocker, just for tracking)
  • beacon-api | produceBlock and publishBlock: Remove blobs and KZG proofs (link)

P1 : Draft Spec PRs

  • beacon-metrics | PeerDAS metrics: add data column, kzg, custody metrics (link)
  • execution-specs | More post Prague re-factors (link)

Spec Updates:

  • Consensus Specs: v1.5.0-beta.4 is scheduled for release Wed April 2nd ( @jtraglia )

Open Discussion

Discussion Item #1: BPO Blob Schedule

Proposal: Blob schedule for Fusaka (here) by @ralexstokes

Step Increase Blob count
T = 0 1x (6, 9)
T = 2 weeks 2x (12, 18)
Previous T + 2 months 2x (24, 36)
Previous T + 2 months 2x (48, 72)

Discussion Item #2: Beacon API getBlobsSidecar Discusussion

summary here

GitHub Issue