[TuP-F] Poster Session 1

[TuP-F-8] Dynamically Retraceable Optical Switching Reconfiguration Scheme for Accelerating Distributed Machine Learning

○Yun Teng¹, Hui Yang¹, Qiuyan Yao¹, Changsheng Yang¹, Zhao Li¹, Jie Zhang¹ (1. Beijing University of Posts and Telecommunications (China))

Keywords:

Switching systems,architectures,and network integrations,Large-scale switching technologies for data center (DC) and high-performance computing (HPC) systems

This paper proposes a reconfigurable configuration scheme for optical interconnect data center networks by introducing the dynamic retraceable algorithm. Simulation results show the proposed scheme significantly reduces packet latency and accelerates distributed machine learning.