Now that this testnet is winding down I’ve been doing something that I held off from earlier because of the churn it will be producing. I’d noticed that the traffic through my router from my 10 nodes had big peaks every 20 seconds or so and wanted to get to the bottom of it and see if it can be eliminated. This is what it looked like:-
That’s the specific switch port the machine with the safenodes uses.
It’s traffic from my nodes out to the internet. ‘Inbound’ on that network port is outbound from the server.
So it’s not much in the grand scheme of things when other householders are working, YouTube’s are being watched etc. but nobody wants big peaks in network traffic if it can be made smoother.
I didn’t think it could be network wide and must be my safenodes.
This is what the traffic from just 1 node looks like (after about 10 mins after the initial flood of chunks from replication has stopped and things have settled down).
The peaks are getting lost a bit in background noise from the server but the peaks are still visible every 20 seconds or so.
So given that nodes seem to go through these convulsions every 20 seconds or so I started my nodes at 3 second intervals. It had previously been my practice to start them at 10 second intervals to give each one a chance to get connected before the next one came along but I think that was longer than necessary.
But I think the main thing if you want to smooth your traffic is start them with an interval that doesn’t align with 20 seconds.
This is what the traffic looks like when 10 nodes have been started at an interval of 3 seconds and the initial flood of replication has stopped.
Much smoother!