Table of Contents

Executive Summary

We analyzed the minimally required duration to execute one HotStuff round for our consensus nodes, comparing data from April 2023 (at this time, we were running a relic-based cryptography stack), Jan 2024 (still running relic-based cryptography stack) and Feb 2024 (after switching to BLST-based cryptography stack). Overall, we are seeing an impressive speedup over time:

1. Between April 2023 and Jan 2024, the time for our implementation to finish the work for one HotStuff round (called view) decreased by about 35% (in absolute terms about 300ms).
2. Tarak's change from relic to BLST (main goal of this switch was to reduce software maintenance efforts and lower the hurdle for 3rd party community contributions) that was rolled out beginning of February resulted in another 60% speedup (300ms) compared to the data one week before.

While our consensus got much faster, we currently don't see a higher block rate or faster time to finality. This is expected, because our cruise control system delays the publication of blocks to maintain a human-specified block rate. We do this to have a reliable Epoch switchover every Wednesday.

Next Steps:

1. Review the results in the Core-Protocol Working Group (next meeting is scheduled March 28)
2. Discussion on how to adjust cruise-control parameters specifically (what values do we pick)
3. Governance multi-sig and possible an update of the Epoch smart contract to deploy the new parameter set for Cruise Control
4. Last step would be to write some technical background content for marketing.

My gut feeling is that we could get to something like a 30 - 50% speedup in time to finality (potentially more depending on our risk tolerance).

Context

Time To Finality [TTF] and Time To Executed [TTE]