Litcius/Paper detail

Astraea: Towards Fair and Efficient Learning-based Congestion Control

Xudong Liao, Han Tian, Chaoliang Zeng, Xinchen Wan, Kai Chen

202426 citationsDOI

Abstract

Recent years have witnessed a plethora of learning-based solutions for congestion control (CC) that demonstrate better performance over traditional TCP schemes. However, they fail to provide consistently good convergence properties, including fairness, fast convergence and stability, due to the mismatch between their objective functions and these properties. Despite being intuitive, integrating these properties into existing learning-based CC is challenging, because: 1) their training environments are designed for the performance optimization of single flow but incapable of cooperative multi-flow optimization, and 2) there is no directly measurable metric to represent these properties into the training objective function.

Topics & Concepts

Convergence (economics)Computer scienceStability (learning theory)Control (management)Network congestionMetric (unit)Function (biology)Flow control (data)Mathematical optimizationDistributed computingArtificial intelligenceMachine learningComputer networkEngineeringMathematicsNetwork packetOperations managementBiologyEvolutionary biologyEconomic growthEconomicsNetwork Traffic and Congestion ControlInternet Traffic Analysis and Secure E-votingSoftware-Defined Networks and 5G