Markdown Version | Session Recording

Session Date/Time: 29 Mar 2023 04:00

iccrg

Summary

This meeting of the Internet Congestion Control Research Group (ICCRG) featured two presentations on congestion control research. The first presentation focused on modeling the interaction between Cubic and BBR congestion control algorithms and predicting the Nash equilibrium distribution of these algorithms on the internet. The second presentation introduced a novel queuing discipline called Deadline Safe Credits (DSC) designed to provide a low-delay forwarding service without degrading best-effort traffic. The meeting also included discussions and Q&A following each presentation.

Key Discussion Points

• Cubic vs. BBR Modeling:
  • Discussion on the assumptions made in the Cubic/BBR model, particularly the assumption of similar RTTs and its impact on the results.
  • The model's ability to predict throughput gains and the Nash equilibrium distribution as more flows switch from Cubic to BBR.
  • Debate on the relevance of the model to real-world scenarios with diverse flow sizes, RTTs, and network conditions, including the presence of middleboxes.
  • Discussion about whether performance-driven congestion control is ever possible.
• Deadline Safe Credits (DSC):
  • The design and implementation of the DSC queuing discipline, focusing on its ability to provide a low-delay forwarding service.
  • The efficiency of the DSC implementation in the Linux kernel and its ability to handle high bandwidth environments.
  • Discussion on the trade-off between packet loss and delay in DSC, and the importance of bandwidth estimation for the credit devaluation scheme.
  • Need for bottleneck routers to implement this capability.
• General Congestion Control Considerations:
  • Discussion on issues related to implementing quality of communication.
  • General comments concerning new coordinated actions in 5G related to internet control.

Decisions and Action Items

• None.

Next Steps

• Continue research on modeling congestion control interactions with diverse network conditions and traffic patterns.
• Explore the implications of different network utility functions beyond throughput in congestion control algorithms.
• Further investigate mechanisms for coordinating actions in 5G and 6G systems to improve network performance and user experience.