Try   HackMD

Geemo Update #4


Hey all! I implemented the beaon apis for light_client_optimistic_update, light_client_finality_update, light_client_bootstrap, and added the light client topics to the eventstream. The PR has been submitted to lighthouse. Additionally I continued exploring what it would take to include tests for the light client rpcs and have a solid understanding of how to achieve it. However the way I see it being done is not very clean and I wanted to get a second opinion, so I contacted @pawan, one of the lighthouse networking gurus, to see if there was a better way that I was missing.

Moving forward the main components left for a working light client server are:

Of course previous PRs also need to be merged as well.

Future Work:

  • caching light_client data for easier access
  • update fetching methods to reflect the experimental tree-states branch (won't be experimental for too much longer)
  • extend the LightClientHeader with the execution payloads to support the capella upgrades to light client data.
  • Add metrics to track the workload of the light client server in practice.

There is still a lot of work to be done, but I'm excited to see it to completion.

This is a running list of PRs submitted to Lighthouse during the EPF program with their merge status listed.

Miscellaneous PRs:

Last changed by 
 
geemo
0
179

Read more

EPF 4 Final Update

This project sought to increase block proposer rewards for Lighthouse users. The project proposal lives here: Lighthouse Attestation Packing.

Nov 6, 2023
EPF Update #10

This update covers the process toward merging the optimality preserving atttestation aggregation process.

Nov 6, 2023
EPF Update 9

This update will cover the current state of the new block packing methods for Lighthouse. With the assistance of Michael Sproul we have an optimality preserving aggregation strategy implemented. We are currently testing and optimizing it’s performance. This comes as a pretty significant change to the block packing logic so we must ensure that everything functions correctly with good performance. Michael set an initial performance goal as attestation packing taking <50ms.

Oct 15, 2023
EPF Update 8

In the past week I tested and optimised the decomposition approach that I described in my previous update. The resulting algorithm foes not perform as well as I would have hoped. That being said it is faster than the base MIP solution in some cases, but it’s not a significant difference. Additionally I’ve tested the aggregation branch and resolved some outstanding issues on it. In the next week I’d like to get the aggregation branch merged or at least looked at by Michael Sproul. It still does not perform as well as the greedy aggregation approach so it certainly isn’t worth merging as is, but I think this could be solved.

Oct 2, 2023
Read more from geemo
Published on HackMD