Quote from: jl777 on February 02, 2014, 07:52:35 AM

Quote from: Come-from-Beyond on February 02, 2014, 07:18:59 AM

Quote from: jl777 on February 02, 2014, 07:16:00 AM

More seriously, is there a way to limit overall network TPS? If so, we can gradually change it to match actual usage levels. If some peak time blocks need to spill over into the following block, that won't be so bad. If 20% of the network gets blacklisted due to lack of bandwidth that seems worse.

Maybe we crank it up to full speed and see what happens? Maybe there is no issue at all?

U could be the guy who answers these questions.

I dont have the data I need to be able to answer any of this. I have no idea if TPS can be capped. If it can, I am assuming it will require a new client version with a new hardcoded cap.

Then again, we might be able to define an acceptable percentage of nodes that are blacklisted due to not enough bandwidth/CPU/etc and have it AUTOMATICALLY adjust the overall network TPS to the acceptable lost nodes percentage. Then if we find that TPS is bottlenecked by too many slow nodes, we upgrade those nodes or turn them into thin clients.

Is this automatic max TPS changing possible to do?

I ** REALLY ** want this dynamic automatic adjusting of overall network TPS. It is like dynamic memory allocation versus static allocation, except we are talking about expensive server costs here.
Managing the NXT network would become much simpler if we just had to identify the lagging nodes and dealing with those, rather than the alternative.

By picking the right percentage of acceptable blacklisted nodes and acceptable percentage of spillover blocks, we could lower overall NXT server costs by 50% or more.

James