some problems in experiments

[ykxian]

Hi @pfandzelter I encountered some problems after repeating the experiment several times.

1. I performed the following operations on the host with Docker installed, and now it seems to be running normally.

   sudo apt-get purge docker-ce docker-ce-cli containerd.io docker-compose-plugin
   sudo apt-get autoremove
   sudo iptables -F
   sudo iptables -t nat -F
   sudo systemctl restart NetworkManager
   sudo systemctl restart systemd-networkd
   sudo rm -rf /var/lib/docker
   sudo rm -rf /etc/docker
   sudo rm -rf /usr/libexec/docker
   sudo rm -rf /var/run/docker.sock
   sudo groupdel docker

   I'm not sure which commands were effective or if they would be helpful to you.

2. The number of packets received is less than expected.

3. The actual latency is very high. After stabilizing, the latency is approximately 4 seconds.

4. When interrupted, this code does not work, causing the subsequent code not to execute. onewaymeasurement.py

I am not sure if questions 2 and 3 are related to my equipment or network environment. Even if I don't use the host with Docker installed, the experiment results are still the same. Thank you!

[pfandzelter]

Hi again, thanks for continuing to try this. I found a bug in Celestial in an unrelated experiment that caused the bandwidth to be only 1/1000th of what it should be. Maybe it's related to this. I will try it out on my setup

[ykxian]

Indeed, it is related to this. I increased the bandwidth by a thousand times, and the results improved significantly.

[ykxian]

[pfandzelter]

That looks much better! I have merged the fixes for Celestial and will shortly update this repository. I have made minor changes to the HTTP API to now reflect the units of information, e.g. delay_us. Not having units caused the previous bug...

[ykxian]

Thanks again! It works well now. The latency fluctuation is somewhat large.

[ykxian]

Sometimes there is negative latency.

[ykxian]

I synchronized the time on three servers, and now it's working well.