IEEE INFOCOM 2020 - IEEE Conference on Computer Communications 2020
DOI: 10.1109/infocom41043.2020.9155250
|View full text |Cite
|
Sign up to set email alerts
|

Eagle: Refining Congestion Control by Learning from the Experts

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
1
1
1
1

Citation Types

0
4
0

Year Published

2021
2021
2024
2024

Publication Types

Select...
5
3

Relationship

0
8

Authors

Journals

citations
Cited by 37 publications
(4 citation statements)
references
References 6 publications
0
4
0
Order By: Relevance
“…The idea is to have two levels of controls: ine-grained control using classic TCP algorithms, e.g., BBR, to adjust the congestion window, and hence the sending rate, of a user, and coarse-grain control using DRL to calculate and enforce a new congestion windows periodically, observing environment statistics. The proposed solution therefore has more predictable performance and better convergence properties, showing how learning from an expert, e.g., BBR algorithm, can improve the performance, in terms of convergence speed, adaptation to newly seen network conditions, and average throughput [41].…”
Section: Ml-based Approachesmentioning
confidence: 93%
“…The idea is to have two levels of controls: ine-grained control using classic TCP algorithms, e.g., BBR, to adjust the congestion window, and hence the sending rate, of a user, and coarse-grain control using DRL to calculate and enforce a new congestion windows periodically, observing environment statistics. The proposed solution therefore has more predictable performance and better convergence properties, showing how learning from an expert, e.g., BBR algorithm, can improve the performance, in terms of convergence speed, adaptation to newly seen network conditions, and average throughput [41].…”
Section: Ml-based Approachesmentioning
confidence: 93%
“…It was estimated that BBR accounts for over 40% of the total traffic on the Internet. [41][42][43][44][45][46][47][48] Remy [41] optimizes and generates congestion control algorithms offline based on pre-specified congestion control objectives, protocol assumptions, and models of both network and traffic. Performance-oriented congestion control (PCC) [42] sends packets at two different rates, slightly higher and slightly lower than the current rate.…”
Section: Survey Of Congestion Control Algorithms Share On the Internetmentioning
confidence: 99%
“…The reward metric can be observed as either the throughput or network delay that is optimised for a particular system state by the actions of the DRL agent using trained DNN [ 112 ]. Some of the CC algorithms that use DRL techniques include Aurora [ 118 ], Eagle [ 119 ] and Orca [ 120 ] ( Table 4 ).…”
Section: Machine Learning For Improving CCmentioning
confidence: 99%