PeerDAS Breakout Room #19

[jimmygchen]

• Feb 25 (Tue), 2025 , 10:00 UTC 🕙 ❗
• Duration: 60 minutes
• Zoom link
• Ethereum Protocol Call calendar
• Notes
• Previous call: PeerDAS Breakout Room #18 #1295

Agenda

• client updates
• devnet updates
• spec discussion
• open discussion

[zilm13]

I want to discuss:
Removing/altering sampling concept from the FULU spec.
After validator custody is added, any validator node requires to have at least 8 columns in custody. So for sampling they will need to check only that custody is filled. Only full nodes (without validator) should store 4 columns in custody, sample (and drop) another 4.

In order to make spec clear and simple and software contains less bugs, I suggest:

• drop sampling concept from F- fork spec. If we need it in G-fork, we could add it in G-fork spec
• is_data_available checks that current (slot, blockRoot) columns custody is completely filled
• CUSTODY_REQUIREMENT constant is bumped from 4 to 8
• SAMPLES_PER_SLOT constant is removed

Only disadvantage: it will require +16Gb of the storage for each 16 blobs/per block for full nodes only, but they will serve at least the same amount of DA as 1 validator node for the network after this change. And any validator node is already required to serve this DA and requires this amount of space.

[nalepae]

drop sampling concept from F- fork spec. If we need it in G-fork, we could add it in G-fork spec
You mean drop peer sampling?

https://github.com/ethereum/consensus-specs/blob/dev/specs/fulu/peer-sampling.md

[jimmygchen]

Summary:

• Client teams are mostly working on validator custody and refactoring and making some progress.
• peerdas-devnet-5 was launched 5 days ago and mostly running smoothly (there's a known grandine issue that needs investigation).
• Teku suggested removing sampling concept from the spec to reduce complexity - teams agree that it adds complexity, but there was no strong opinions on changing the spec as most teams have already implemented this. Teams to discuss offline (we should finalise this ASAP though)
• Nimbus raised a point about researching / estimating network structure with validator custody in place, in order to know more about potential sync speed on mainnet. Will add an R&D task to the readiness checklist.
• Breakout call will be back to 1400 UTC next tuesday!

Notes here: https://docs.google.com/document/d/1Ng2IrCe28kTt1BnIsjtMlKHq2MHgaja24LhFXSvqfJQ/edit?usp=sharing

[jimmygchen]

Next call #1326