Ah, I see. you mean for a download task, the upstream traffic is higher than the downstream traffic.
I can’t explain it either.
One guess I have is that:
libp2p using QueryIterator to get_closest, which includes dialing and sending out identify messaging to nodes along the path
due to the super large size of the network, the hop involves in the Query becomes more and more
there is also network discovery involves extra calls of get_closest queries as well
such accumulated dial/identify messaing could exceeds the size of the data to be downloaded itself.
This is just my guess. May need further work to confirm or invalidate it.
Yea, I was expecting it to be a lot less. I have limited uplink b/w so have to be very careful with running nodes due to that. 50 download instances running is using more uplink b/w than 1250 nodes by 20%
This could be an issue in future for efficiency reasons and having multiple people using Autonomi behind the one router on average ISP connection ( world average < 40Mbps uplink for most ISP connections ). For instance a youtube replacement application being used by a few people is mostly downloading a lot of data and if upload b/w is higher then this limits them a lot.
Anyhow I will leave it in your capable hands to examine this when there is time down the track