The program committee has accepted a BoF session.
Cellular networks exhibit significant variations in bottleneck link rates even over short time durations. AQM techniques like fq-codel with ECN feedback are insufficient.
It is challenging to properly evaluate congestion control approaches over cellular network paths; it gets difficult when it involves in-network mechanisms, since it is not easy to deploy a test network running new network-layer mechanisms in such setups.
Hari Balakrishnan et al will lead a BoF discussion to help formalize the constraints of the problem and formulate ways in which the community should evaluate congestion control protocols for cellular networks.
This BOF will put forward an evaluation technique involving collection of real-world cellular network traces and then replaying them on Linux based infrastructure. The Mahimahi toolset facilitates such a setup utilizing Linux containers. This approach allows for experimenting and validating newly coded algorithms on Linux in an accurate environment.
To kick things off in this BoF and frame future discussions, Hari et will illustrate how to use collected traces and evaluate the Accel-Brake Control (ABC) approach using the Mahimahi toolset.
More info at: https://www.netdevconf.org/0x12/session.html?congestion-control-for-cellular...
cheers, jamal