Search for: All records

The Internet has become the central source of information and communication in modern society. Congestion control algorithms (CCAs) are critical for the stability of the Internet: ensuring that users can fairly and efficiently share the network. Over the past 30 years, researchers and Internet content providers have proposed and deployed dozens of new CCAs designed to keep up with the growing demands of faster networks, diverse applications, and mobile users. Without tools to understand this growing heterogeneity in CCAs deployed on the Internet, the fairness of the Internet is at stake. Towards understanding this growing heterogeneity, we develop CCAnalyzer, a tool to determine what CCA a particular web service deploys, outperforming previous classifiers in accuracy and efficiency. With CCAnalyzer, we show that new CCAs, both known and unknown, have widespread deployment on the Internet today, including a recently proposed CCA by Google: BBRv1. Next, we develop the first model of BBRv1, and prove BBRv1 can be very unfair to legacy loss-based CCAs, an alarming finding given the prolific deployment of BBRv1. Consequently, we argue the need for a better methodology for determining if a new CCA is safe to deploy on the Internet today. We describe how the typical methodology testing for equal-rate fairness (every user gets the same bandwidth) is both an unachievable goal and ultimately, not the right threshold for determining if a new CCA is safe to deploy alongside others. Instead of equal-rate fairness, we propose a new metric we call, harm, and argue for a harm-based threshold. Lastly, we present RayGen, a novel framework for evaluating interactions between heterogeneous CCAs. RayGen uses a genetic algorithm to efficiently explore the large state space of possible workloads and network settings when two CCAs compete. With a small budget of experiments, RayGen finds more harmful scenarios than a parameter sweep and random search. 

The rise of proprietary and novel congestion control algorithms (CCAs) opens questions about the future of Internet utilization, latency, and fairness. However, fully analyzing how novel CCAs impact these properties requires understanding the inner workings of these algorithms. We thus aim to reverse-engineer deployed CCAs' behavior from collected packet traces to facilitate analyzing them. We present Abagnale, a program synthesis pipeline that helps users automate the reverse-engineering task. Using Abagnale, we discover simple expressions capturing the behavior of 9 of the 16 CCAs distributed with the Linux kernel and analyze 7 CCAs from a graduate networking course. 

Much of our understanding of congestion control algorithm (CCA) throughput and fairness is derived from models and measurements that (implicitly) assume congestion occurs in the last mile. That is, these studies evaluated CCAs in “small scale” edge settings at the scale of tens of flows and up to a few hundred Mbps bandwidths. However, recent measurements show that congestion can also occur at the core of the Internet on inter-provider links, where thousands of flows share high bandwidth links. Hence, a natural question is: Does our understanding of CCA throughput and fairness continue to hold at the scale found in the core of the Internet, with 1000s of flows and Gbps bandwidths? Our preliminary experimental study finds that some expectations derived in the edge setting do not hold at scale. For example, using loss rate as a parameter to the Mathis model to estimate TCP NewReno throughput works well in edge settings, but does not provide accurate throughput estimates when thousands of flows compete at high bandwidths. In addition, BBR – which achieves good fairness at the edge when competing solely with other BBR flows – can become very unfair to other BBR flows at the scale of the core of the Internet. In this paper, we discuss these results and others, as well as key implications for future CCA analysis and evaluation. 

The Internet community faces an explosion in new congestion control algorithms such as Copa, Sprout, PCC, and BBR. In this paper, we discuss considerations for deploying new algorithms on the Internet. While past efforts have focused on achieving 'fairness' or 'friendliness' between new algorithms and deployed algorithms, we instead advocate for an approach centered on quantifying and limiting harm caused by the new algorithm on the status quo. We argue that a harm-based approach is more practical, more future proof, and handles a wider range of quality metrics than traditional notions of fairness and friendliness. 

BBR is a new congestion control algorithm (CCA) deployed for Chromium QUIC and the Linux kernel. As the default CCA for YouTube (which commands 11+% of Internet traffic), BBR has rapidly become a major player in Internet congestion control. BBR’s fairness or friendliness to other connections has recently come under scrutiny as measurements from multiple research groups have shown undesirable outcomes when BBR competes with traditional CCAs. One such outcome is a fixed, 40% proportion of link capacity consumed by a single BBR flow when competing with as many as 16 loss-based algorithms like Cubic or Reno. In this short paper, we provide the first model capturing BBR’s behavior in competition with loss-based CCAs. Our model is coupled with practical experiments to validate its implications. The key lesson is this: under competition, BBR becomes window-limited by its ‘in-flight cap’ which then determines BBR’s bandwidth consumption. By modeling the value of BBR’s in-flight cap under varying network conditions, we can predict BBR’s throughput when competing against Cubic flows with a median error of 5%, and against Reno with a median of 8%.