Curious. I’ll run some more tests on my boxes.
With upnp enabled (but struggling to function), I was getting a traditional spread of around 125MB resident per node. With home network, I’m getting about half that.
I’m building a near identical box later, so will try running them side by side, one with port forwarding, one with home network, then check results.
What’s the best way to calculate the average node memory usage?
Run 10 nodes and see the increase in memory usage
Thinking about this 25% lower HOME+STARLINK, , any Satellite connection has more tail latency and higher packet drop and resend counts, so this is likely the root cause of the 15% less earn ‘settled’ rate…, in that its possible some nodes were shunned for a period of time, then rejoined (ie as I understand it three strikes your out if an antnode is not responsive to connection request, shunning policy?).
The combination of physical terrestrial distance over earth surface communications to the corresponding connecting node (making the request) would likely have greater distance and therefore greater amount of link latency delay between router relays, etc…, for this to happen,
and it will happen, given Random XOR node address generation and no geographic zoning (regional caching) to optimize node2node interconnect speeds currently(to reduce the chance of shunning caused by WAN latency due to higher number of relays in the node2node path).
Actually the time to the USA over starlink is shorter than for the HFC link. Lower latency for the most part and with the laser comms + more satellites coming in it should get lower. And much faster to the EU.
Relay nodes are now going to only be used to establish the connection, after that there is no need of the relay node and comms will be as fast as port-forward until the connection drops.
Shunning is not based on latency but non-communications and failing storage checks. Also shunning is permanent and short time shunning is not a thing.
And I have to come back to doing tests as I have other tasks that i need to get done, like finish off the bookshelves/display cabinet. Been doing it part time and modding things, waiting to get needed parts, and so on has made the living room a war zone and enough is enough, so they are running with 3 SBC using port forward and one using home-network. All earning something and with the network so large now its rather random even on the Big machine over the HFC link.
With that randomness and slowness of earning for any node now, any testing will have to take some time and ensure the machines are kept running.
Do you know when that’s expected to land?
Nope no info, depends on testing