Contention Detection on the Internet
The current heterogeneity in the Internet’s congestion control landscape [1] has prompted a lot of discussion around fairness, and how congestion control algorithms need stricter deployability standards in order for the Internet to continue to work safely. However, there exists a camp [2] that argues fairness might not be as critical as we think it is for the Internet to work. Their running hypothesis is that given the current economic arrangements around Internet access, most flows might already be isolated from each other on the Internet - thereby making fairness concerns moot.
The goal of this thesis is to answer a simple question: Do flows compete for bandwidth with other flows on the Internet? The student will have to combine and extend existing active measurement techniques, as well as build a custom congestion control algorithm that can detect contention over paths on the Internet. This measurement study holds high consequence for Internet congestion control research over all, and is likely to ultimately contribute to an academic publication.
Requirements
- Good understanding of computer networking and congestion control
- Strong programmin skills, preferably in Rust or C
References
- Mishra et al. “Keeping an Eye on Congestion Control in the Wild with Nebby” Proceedings of SIGCOMM 2024
- Brown et al. “How I Learned to Stop Worrying About CCA Contention” Proccedings of HotNets 2023
- Goyal et al. “Elasticity Detection: a building block for Internet congestion control” Proceedings of SIGCOMM 2022
Supervisors
