This looks somewhat odd, the first and last values imply the measurement might not be have been long enough.
weird. I didn't stop the script, it ended by itself.
I'll try again.
I've done some tests : it seems better. Updating a game on a PC a watching a movie on the other PC (always through SSH mount) makes it less glittering (or not at all with 480p movies).
I don't think we can do better with such an connection. I'll try to look for a video player that can buffer more (didn't find it onVLC)
Here is another result of ATM overhead detector, that ran for less than hour and stopped 
Thanks for your time and explanations @moeller0 . That's better now 
Mmmh...
Would you mind sharing the data file with me, so I could have a look at what is going on here? My guess is that the ICMP reflector only responds to a small set of ICMP packet sizes and that causes some issues... which IP address did you use as reflector?
I didn't change the reflector and kept the one in the script : Google DNS 8.8.8.8
Here are the ping results : https://filetransfer.io/data-package/OzCochKm#link
Ah, it turns out 8.8.8.8 will not reflect arbitrarily sized ICMP echo requests, but limits thngs
´´´
user@123-1234563 space % ping -c 2 -s 95 8.8.8.8
PING 8.8.8.8 (8.8.8.8): 95 data bytes
76 bytes from 8.8.8.8: icmp_seq=0 ttl=115 time=10.822 ms
wrong total length 96 instead of 123
76 bytes from 8.8.8.8: icmp_seq=1 ttl=115 time=10.974 ms
wrong total length 96 instead of 123
--- 8.8.8.8 ping statistics ---
2 packets transmitted, 2 packets received, 0.0% packet loss
round-trip min/avg/max/stddev = 10.822/10.898/10.974/0.076 ms
´´´
Please try 1.1.1.1 instead of 8.8.8.8 as it seems 1.1.1.1 still responds to larger packets...
I do believe though that the result might actually be correct, but with less than 2 cell quanta in the plot that is just that a believe/hope that should be confirmed...
Same thing with 1.1.1.1
File is same size than with 8.8.8.8 : 8 MB
Here it is : https://filetransfer.io/data-package/kq1AWL41#link
So switching from median to truncated_mean (robust_mean in the code) we get*:
Now we see at least the expected number of samples. However the number 43 is completely new to me, I would recommend sticking to the 44 you have right now.
*) Probably required due to the level of noise in the data...
1 Like
Got it !
Thanks again for your time and analysis.
Have a great day 