Rethinking Congestion Control in 5G

Overview

The shift to 5G networks has introduced a wave of network challenges—short-timescale variability, high-throughput demands, and bottleneck buffer sensitivity. Two novel frameworks, ALCC and Zeus, offer orthogonal but complementary approaches to tackling these issues. ALCC enables user-space deployment of congestion control over legacy TCP tunnels, making real-world deployment feasible even with application-layer constraints. Zeus presents a unified benchmarking framework to rigorously evaluate congestion control algorithms in real and emulated 5G environments. Together, they pave the way for a rethinking of congestion control strategies in mobile systems.

Key Capabilities

• Application-Layer Congestion Control over TCP tunnels without OS/kernel changes
• Easily deploy Verus, Sprout, and Copa in user-space with comparable performance
• Unified 5G CCA benchmarking framework supporting real and emulated network traces
• Enhanced Mahimahi for multi-Gbps 5G channel emulation
• Side-by-side comparison of 10 CCAs with fairness and harm analysis

Results

• Achieves 90–97% of native CC performance (throughput & delay) from user-space
• Enables server-only deployment for Verus and Copa with zero client-side changes
• Cubic and Reno dominate throughput on large-buffer 5G channels; Copa and Vivace excel at low latency
• BBR is the most balanced across throughput and delay tradeoffs in varied 5G topologies
• Includes fairness and harm metrics to evaluate CCA coexistence with legacy TCP

Papers

• ZEUS: An Experimental Toolkit for Evaluating Congestion Control Algorithms in 5G Environments
• ALCC: Migrating Congestion Control to the Application Layer in Cellular Networks

Collaborators

• Muhammad Khan (NYUAD)
• Rohail Asim (NYU)
• Luis Diez (University of Cantabria)
• Ramon Aguero (University of Cantabria)
• Lakshmi Subramanian (NYU)
• Yasir Zaki (NYU Abu Dhabi)