ethersphere
Define the max value of maxAllowedDoubling where performance is still good
We need to perform some tests in order to understand what is the maximum height that would be still performant for the node and the network.
I suggest the following to be taken into account:
- the node needs to be "fully synced" and staked; the other variables need to be sufficient for it to play: enough funds for gas, stable RPC endpoint
- try out different doubling variants and test at which "height" the node no longer plays succesfully; and make sure the behaviour is consistent (e.g. not just one win, which could be due to random factors aligning)
- the test should probably first be done on testnet, and also some traffic simulated to increase load on the machine(s)
- the test should probably also be done on mainnet, to cancel out any influences of it (e.g., the calculation time on testnet using Sepolia is much longer than mainnet, due to different block times?)
Really also research should give inputs, as I think they know more where the issue might be with higher doublings.
From my personal experience, I could run a node with 2 ^ 6 = 64 doubling. That one had trouble staying in sync. Lowering the doubling made it possible to stay in sync on paper, but it still would not play. Lowering a couple more times it actually played, but I think it did not commit a hash (not sure), though it did seem to calculate it. Not sure about this, but the point is, probably only winning would demonstrate that a doubling level is appropriate.
(Not to mention that there is currently very little traffic on mainnet, so the conditions are as good as they will get.)
