One would expect some correlation between bandwidth usage and air traffic. I cannot see much of that here. The first picture shows bandwidth usage this morning and the second aircraft seen. (There is a time difference, but look at the common parts.)
A friend of mine sees essentially the same traffic, although I usually see a few more a/c. His bandwidth usage seems to be more in line with what you expect:
We both have MLAT working.
Can someone explain this?
I doubt that you are measuring what you think you are measuring.
Typical upload rates for piaware are <10kB/s including mlat. Neither set of graphs agrees with that.
You’re probably including skyaware traffic (which is purely local), and maybe something else that’s causing the big spikes in the first set of graphs.
At a time I had nothing but a Pi on my network for a month, the modem recorded 20Mb/day without MLAT and 60MB/day when MLAT was enabled.
I see. I should have realized that the graph includes local traffic. Have to think about what my friend and I are doing differently. Also, I often see more bandwidth usage here in the afternoons than in the mornings but I don’t think my habits differ.
When I compare the number of aircraft seen and the number of positions reported different days I also see that even if the the number of a/c is essentially the same, the number of positions might differ by 25 % or more. So I started to suspect that there might be a communication problem. Using mtr I do see packet loss on the way to FA. Could this be the reason for the differences?
I just saw that during the time I wrote my first reply the bw usage went up to about 40 kbyte/s continous throughput. And so far I have not done anything which could cause local traffic. (If the web page is not updated in the background?)
Well, except that I looked at the graph of course!
That is not much! I think an average of 0.7 kbyte/s in the MLAT case.
MLAT setting forces the piaware to send data more often to the servers, to maintain the sync. All data that is.
If one wishes to determine what network activity is occurring, I suggest using a network monitoring tool such as PRTG (easy) or MRTG (harder to configure).
The graphs auto refresh which uses more than no bandwidth.
Might well be the cause 
After all that’s using local bandwidth.
I see, but I am still confused. The graph below is typical for the afternoons and you have seen that the mornings are different. I don’t think that my habits differ that much during the day.
Well disable lighttpd for a day then.
sudo systemctl disable --now lighttpd
After a day enable it again.
sudo systemctl enable --now lighttpd
That way you know if it’s you or someone else having some page open.
Yes, I could do that but now, thanks to this discussion I am quite confident that it is local traffic:
If you use the local Map displayed locally, that’s also counted as traffic. But it doesn’t leave your local network.
I am feeding several sites, this causes approx 1 GB of traffic outbound per day. But the majority is caused by opensky-network which has a different approach. Without that it’s significant lower.
For the total amount leaving my local network i have darkstats running on my Raspberry. It can list the traffic per host and you can get a good overview what causes the most traffic.
Darkstat - A Web Based Linux Network Traffic Analyzer (tecmint.com)
I didn’t know about darkstat. Sounds like a perfect tool to check what causes the traffic.
I also found a lot of interesting statistics in my router.
Now, the amount of traffic is no problem. I have a high capacity fiber connection. I was more concerned about that the feeder was not working correctly.
Anyway I guess FA does not fetch all data from all feeders as that would mean a lot of duplication in areas with many feeders. Still, they would like to find out if a feeder has unique info. Would be interesting to know how they handle that.
A different solution would be NTopNG, but that might be way too much for your topic.
The graphs simply do not distinguish the traffic between external and internal. Normally i don’t care about these graphs, there are others more interesting, e.g. the I/O which should be as low as possible to avoid getting the SD Card killed too fast.
Read this code: https://github.com/flightaware/dump1090/blob/master/net_io.c#L2552
Anyhow … at certain intervals info about an aircraft is sent to FA …
Deduplication is server side.
If you only send info about an aircraft every 15 seconds … it’s not that much traffic if you consider compression.
Thanks! Darkstat installed and working great. Will look at ntopng later.
Thanks for the tip! I will have a look.
As said, might be an overload, like getting a 40 ton truck for just getting a fridge from the store to your home…
If you’re fine with darkstat - stay with it.