stephenakridge
186 posts
Central Scheduler (in internal tests) showed 80% higher fee collection. This is very different than what it might be in production. So let’s see how that goes .
@ctjlewis Shouldn’t be insane, problem is they changed the way they measure capacity usage from actual mem and core usage to the new “CU” which is just abstracted mem and core usage. And after the change some of our pipelines have been fucked since and they can’t explain why
Just read this whole thing end-to-end touches on DePIN, Firedancer, Tinydancer, Jito, Compression, DeFi and misconceptions on hardware requirements and more — really well done! here's the concluding paragraph, the fundamentals haven't changed, the future remains bright
leader still needs to maintain state but they cant execute before propagating, you can put older state in the current block cos execution is trailing consensus / block production, the info can be stored as a sep tree in the bankhash but again there needs to be a concrete proposal for this design because there are diff ways it can be implemented and can introduce major changes which is why only labs can decide here, even the simd hasnt be touched since december
