An experimental study of the learnability of congestion control

By Anirudh Sivaraman, Keith Winstein, Pratiksha Thaker, Hari Balakrishnan. 2014

137 citations recorded by the source corpus.

When designing a distributed network protocol, typically it is infeasible to fully define the target network where the protocol is intended to be used. It is therefore natural to ask: How faithfully do protocol designers really need to understand the networks they design for? What are the important signals that endpoints should listen to? How can researchers gain confidence that systems that work well on well-characterized test networks during development will also perform adequately on real networks that are inevitably more complex, or future networks yet to be developed? Is there a tradeoff between the performance of a protocol and the breadth of its intended operating range of networks? What is the cost of playing fairly with cross-traffic that is governed by another protocol? We examine these questions quantitatively in the context of congestion control, by using an automated protocol-design tool to approximate the best possible congestion-control scheme given imperfect prior knowledge about the network. We found only weak evidence of a tradeoff between operating range in link speeds and performance, even when the operating range was extended to cover a thousand-fold range of link speeds. We found that it may be acceptable to simplify some characteristics of the network---such as its topology---when modeling for design purposes. Some other features, such as the degree of multiplexing and the aggressiveness of contending endpoints, are important to capture in a model.

Subfields: Computer Networks and Communications

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Networks and Communications

Citations in the corpus

Certified Defenses against Adversarial Examples (2018, background): “More broadly, certiﬁcates of performance are likely necessary for deploying machine learning systems in critical infrastructure such as internet packet routing (Winstein & Balakrishnan, 2013; Sivaraman et al., 2014).”
Learning in situ: a randomized experiment in video streaming (2019, background): “Machine-learned systems in computer networking sometimes describe themselves as achieving near-“optimal” performance, based on results in a contained or modeled version of the problem [25, 37, 39].”
Congestion Control for Large-Scale RDMA Deployments (2015, background): “Recent proposals consider the possibility in leveraging machine learning to adapt TCP parameters for better performance [36, 40].”
PCC Vivace: Online-Learning Congestion Control (2018, background): “While they can attain high performance, Remy-generated TCPs are inherently prone to degraded performance when the actual network conditions deviate from input assumptions [27].”
Copa: Practical Delay-Based Congestion Control for the Internet (2018, background): “A fourth, and most recent, thread of end-to-end congestion control research has argued that the space of congestion control signals and actions is too complicated for human engineering, and that algorithms can produce better actions than humans (e.g., Remy [37, 43], PCC [9], and PCC-Vivace [10]).”