Understanding BBRv3 Performance in AQM-Enabled WiFi Networks
Shyam Kumar Shrestha, Jonathan Kua and Shiva Raj Pokhrel · 2025
Read Understanding BBRv3 Performance in AQM-Enabled WiFi Networks by Shyam Kumar Shrestha, Jonathan Kua and Shiva Raj Pokhrel (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2509-06245.txt.
Introduction
Google's newest congestion control algorithm behaves differently on managed home Wi-Fi, and Shyam Kumar Shrestha and two co-authors measure it in a 2025 paper. Their modular testbed pairs a commercial router with a live dashboard for throughput, latency, and fairness. BBRv3, which paces traffic by bottleneck bandwidth and round-trip time, faces the loss-based CUBIC under three queue management schemes.
BBRv3 improves fairness and convergence notably under active queue management, especially FQ-CoDel. The testbed and visualization tool are offered as a practical base for judging next-generation transport variants on real wireless links rather than simulations alone.
Catalog record
| CREATOR | Shyam Kumar Shrestha, Jonathan Kua and Shiva Raj Pokhrel |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 17 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2509-06245
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Understanding BBRv3 Performance in AQM-Enabled WiFi Networks Shyam Kumar Shrestha, Jonathan Kua and Shiva Raj Pokhrel Abstract We present a modular experimental testbed and lightweight visualization tool for evaluating TCP congestion control performance in wireless networks. We compare Google’s latest Bottleneck Bandwidth and Round-trip time version 3 (BBRv3) algorithm with loss-based CUBIC under varying Active Queue Management (AQM) schemes, namely PFIFO, FQ-CoDel, and CAKE, on a Wi-Fi link using a commercial MikroTik router. Our real-time dashboard visualizes metrics such as throughput, latency, and fairness across competing flows. Results show that BBRv3 significantly improves fairness and convergence under AQM, especially with FQ-CoDel. Our visualization tool and modular testbed provide a practical foundation for evaluating next-generation TCP variants in real-world AQM-enabled home wireless networks. I Introduction The widespread use of Wi-Fi networks for data transmission, particularly in residential and enterprise environments, introduces significant complexities in Transmission Control Protocol (TCP) performance. Key challenges include achieving higher resource utilization, maintaining minimum latency maximizing throughput [ 1 , 2 ] , and minimizing packet loss and retransmissions. In response to these ongoing issues, researchers and developers have proposed improvements to TCP’s inherent congestion control mechanisms. Over time, Congestion Control Algorithms (CCAs) have evolved to address the performance demands of increasingly heterogeneous and dynamic network environments. A recent milestone in congestion control evolution is the development of Bottleneck Bandwidth and Round-trip time version 3 (BBRv3) [ 3 ] , developed and deployed on Google platforms since 2023. Building on lessons learnt from BBRv1 [ 4 ] and BBRv2 [ 5 ] , BBRv3 enhances fairness with competing flows, especially in shallow-buffer and dynamic network conditions. It addresses key limitations in earlier versions by refining pacing, congestion window control, and responsiveness to loss signals. Notably, BBRv3 improves intra-protocol fairness and ensures more consistent bandwidth probing, helping flows converge more equitably in both low- and high-buffer environments. In this demo, we present a modular experimental testbed with a visualization tool to highlight BBRv3’s improved performance in contention-limited Wi-Fi setups, when competing with loss-based CCAs such as TCP CUBIC [ 6 ] . Our work serves as a practical tool for evaluating network performance for new and emerging CCAs in a home Wi-Fi networks [ 7 , 8 ] . Our core observation revolves around whether BBRv3 provides demonstrable advantages and is indeed better suited for a simple home Wi-Fi environment than widely-deployed loss-based CCAs, such as TCP CUBIC. Furthermore, we evaluate their actual performance under an emulated 10 Mbps bottleneck, a common constraint in many home internet connections. We test these TCP algorithms with different queuing disciplines (qdisc), ranging from the fundamental Packet First-In First-Out (PFIFO) to more sophisticated approaches such as FlowQueue-Controlled Delay (FQ-CoDel) [ 9 ] and Common Applications Kept Enhanced (CAKE) [ 10 ] . This multi-faceted analysis will provide deeper insights into how each algorithm manages congestion and maintains data flow stability across diverse queuing strategies in practical home settings and Internet of Things (IoT) environments [ 11 , 12 ] . II Data Visualization To visualize the dynamics of competing TCP flows in real-time, we built a lightweight Python-based dashboard using Plotly’s Dash 1 1 1 https://dash.plotly.com/ framework. This dashboard enables clear, customizable visual comparisons of transport-layer performance, especially useful for showcasing how modern congestion control mechanisms such as BBRv3 react to congestion, compete for bandwidth, or adapt to varying wireless link dynamics. (a) Throughput comparison between BBRv3 and CUBIC in competing upload scenario over CAKE AQM. (b) Pacing behavior of BBRv3 when competing with CUBIC over CAKE AQM. (c) Pacing gain dynamics of BBRv3 when competing with CUBIC over CAKE AQM. Fig. 1 : Dashboard of our data analysis and visualization tool. Fig. 1 provides a representative example of the interactive visualization tool developed for analyzing TCP congestion control algorithm performance within our experimental Wi-Fi testbed. This tool facilitates the detailed exploration of key TCP metrics, including throughput, RTT, jitter, and congestion window ( cwnd ), retransmissions for competing flows (e.g., CUBIC vs. BBRv1, BBRv2, BBRv3) under different AQM schemes (PFIFO, FQ-CoDel, CAKE). BBRv3-specific parameters such as pacing gain and pacing rate are also included. For instance, Fig. 1(a) illustrates throughput performance when BBRv3 competes with CUBIC, showing clear disparities under PFIFO. Fig. 1(b) highlights BBRv3’s pacing rate adaptation in such conditions, while Fig. 1(c) depicts its pacing gain across phases and its link to rate changes and throughput variability. By offering an interactive selection of TCP CCAs, AQMs, and performance metrics, the dashboard enables real-time comparison and deep dives into the dynamic behavior of these algorithms across a bottleneck with configurable bandwidth and latency requirements. Furthermore, the modular design of our experimental setup and data analysis/visualization framework allows for ready adaptation to assess new CCAs, diverse network topologies, or varying bottleneck capacities in future research. Visual representation and analysis of CCAs behaviors are crucial for understanding how they interact, adapt and perform in practical Wi-Fi environments. III Modular Experimental Testbed This section outlines our modular experimental testbed setup, which is used in conjunction with our visualization framework for real-time evaluation of CCAs performance in Wi-Fi networks. Fig. 2 : Modular experimental testbed for evaluating TCP BBRv3 under controlled AQM-enabled Wi-Fi bottlenecks. Fig. 2 illustrates a customized wireless testbed architecture designed to evaluate CCAs under realistic Wi-Fi conditions, encompassing both homogeneous and heterogeneous scenarios. At the core is a MikroTik router (hap ax3, RouterOS v7.8; SSID: Mikrotik2), configured with two independent wireless interfaces and distinct subnets to emulate upload and download bottlenecks separately. The sender nodes (Ubuntu 22.04.5 LTS) are connected to ‘wifi1‘, which is bridged with the Ethernet interface (‘Bridge-LAN‘) and assigned to the 192.168.10.0/24 subnet. The receiver node connects to a separate Wi-Fi interface, ‘wifi2‘, operating within the 192.168.20.0/24 subnet. This network segmentation facilitates precise bandwidth shaping and queuing control through MikroTik’s Queue Tree and Firewall Mangle rules. It ensures that queue management policies can be applied independently in each direction, enabling accurate emulation of congestion, contention, and queuing behaviors for robust evaluation of CCAs under Wi-Fi impairments. To emulate constrained wireless bandwidth, MikroTik’s Simple Queue and Queue Tree mechanisms are configured to limit throughput to 10 Mbps in both the upload and download directions. AQM schemes are configured using a strict PFIFO policy with a fixed buffer size of 50 packets, alongside modern AQM schemes, such as FQ-CoDel and CAKE, to enable a comparative analysis of PFIFO vs AQM schemes. This setup allows the study of congestion buildup, tail-drop behavior, and fairness among different CCAs, including BBR variants and CUBIC, under contention-limited Wi-Fi conditions. TCP bulk transfer traffic is generated using iperf3 (v3.9), while real-time socket-level statistics are collected via ss -tin . Packet-level traces captured using tshark (v4.4.6). The raw outputs are parsed into structured JSON logs using the cJSON library (v1.7.3). Additional client-side metrics such as throughput, cwnd , RTT, and jitter are simultaneously logged. The generated experimental datasets support modular and reproducible evaluations of TCP behaviors under Wi-Fi-specific conditions, such as channel contention, bursty losses, and queue-induced bottlenecks. We have made our experimental scripts and detailed router configuration/host setup instructions available via our GitHub repository. 2 2 2 https://github.com/Shyamesh/LCN_Demo . Fig. 3 : AQM configurations in MikroTik RouterOS (hAP ax3, version 7.8), used in our experimental