Work

Load-Aware Ring All-Reduce Routing and Resource Allocation for Distributed AI Training in Artificial Intelligence Data Centers (AIDC)

Year: 2026

Type: conference-paper

Abstract: This study proposes a load-aware Ring All-Reduce algorithm for distributed AI training in AIDCs, jointly optimizing compute-node selection and communication paths using network congestion and GPU states, thereby reducing blocking probability and improving load balancing.

Source: International Conference on Optical Communications and Networks (ICOCN)

DOI: 10.1109/icocn70692.2026.11656042

Authors: Chenyu Zhao, Xin Li, Tianhao Liu, Dongrou Wang, Xunhao Yan, Mingming Sun, Wensheng Zhai, Shanguo Huang

Institutions: Beijing University of Posts and Telecommunications, North China University of Water Resources and Electric Power

Language: English

Topic: Cloud Computing and Resource Management

Subfield: Information Systems

Field: Computer Science

Domain: Physical Sciences

SDG: Industry, innovation and infrastructure

Open Access status: hybrid

OpenAlex ID: W7203825928

FWCI:

Cites: 9

Cited by:

Related to: 10