@Dusk third node type. finally got round to the prover node page and its shorter than i expected 😂
provisioners do consensus. archive nodes hold history. prover nodes generate PROOFS.$BTW
and the reason thats a separate machine is the interesting bit. proof generation is computationally expensive in a way that block production isnt, different hardware profile entirely, so bundling it into the consensus role would mean every provisioner needs to be sized for the heaviest job on the network.
separating it means you can run consensus on modest hardware and let proving happen where the compute actually is.
which connects to something i wrote about earlier in the week, the under-2-second in-browser proving in Hedger. those are two different proving contexts. client side, small circuits, fast. and network side, heavier, dedicated machines. $ACE
so proving isnt one thing here, its a spectrum, and where it runs depends on what its proving.
i still dont know what the actual hardware requirments are, the page didnt say or i missed it.
#dusk @Dusk $DUSK
provisioners do consensus. archive nodes hold history. prover nodes generate PROOFS.$BTW
and the reason thats a separate machine is the interesting bit. proof generation is computationally expensive in a way that block production isnt, different hardware profile entirely, so bundling it into the consensus role would mean every provisioner needs to be sized for the heaviest job on the network.
separating it means you can run consensus on modest hardware and let proving happen where the compute actually is.
which connects to something i wrote about earlier in the week, the under-2-second in-browser proving in Hedger. those are two different proving contexts. client side, small circuits, fast. and network side, heavier, dedicated machines. $ACE
so proving isnt one thing here, its a spectrum, and where it runs depends on what its proving.
i still dont know what the actual hardware requirments are, the page didnt say or i missed it.
#dusk @Dusk $DUSK
