Question about configuring network traces with bandwidth information

[jialianchen]

Thank you for your open source work which has been very helpful for newcomers in this research area!

I have some questions about extending the system to work with new network traces similar to classic datasets like FCC and Norway. I want to test with custom network traces but I'm unclear about where to make the modifications.

In docker-compose-emulation.yaml, I only found:

nginx:
  image: clarkzjw/dashjs-nginx:emulation
  build:
    context: .
    dockerfile: Dockerfile-nginx-emulation
  container_name: dashjs-nginx
  volumes:
    - ./runner/trace/isl-trace.csv:/trace/trace.csv

The isl-trace.csv only contains columns for "since", "relative_seconds", and "rtt", but no bandwidth settings. Could you please clarify:

1. Where is the bandwidth trace configured in the system?
2. What's the proper way to add custom network traces with bandwidth information?

Thank you very much for your help!

[clarkzjw]

Hi @jialianchen, thanks for the interest.

In this implementation, we focused on investigating what is the impact of latency fluctuation to video streaming, and we set a fixed bandwidth setting in all scenarios and let congestion control and ABR algorithms to adapt the latency fluctuations.

If bandwidth traces are needed, an intuitive approach is to modify tc-trace.py and add custom tc command to modify the bandwidth setting of the links.

A better approach might be replacing the original shaper implementation, with custom approaches such as eBPF-based bandwidth shaper or others. You do not necessarily need to implement them in the tc-trace.py. You can implement them independently, and add them to the Nginx container and start with supervisord.

Other approaches include integrating mahimahi, etc.