苏州科技大学学报（自然科学版）2026，Vol.43Issue(1)：72-80,9.DOI:10.12084/j.issn.2096-3289.2026.01.010
高可扩展性数据中心多对一通信拥塞控制方法
A multi-to-one communication congestion mitigation method for highly scalable data center networks
摘要
Abstract
In scenarios where numerous senders communicate with a single receiver,data is consolidated from vari-ous origins into one recipient.The receiver's handling of multiple data streams cannot swiftly complete the communi-cation process,leading to network congestion and severe problems that negatively impact the data center's perfor-mance and scalability.To ensure prompt responses and manage network congestion efficiently,this paper proposes a multi-to-one communication congestion control method tailored for highly scalable data center.By examining the congestion mechanisms in data center's many to one communication,the priority of data packets was established based on their type,originating node,and the geographic location of the forwarding node,thus determining the se-quence of congestion control.This enhanced the precision of congestion management and the overall efficiency of communication.Real-time monitoring of buffer queue occupancy dynamically assesses network communication con-gestion.Upon detecting congestion,a probability function and Lagrange multiplier technique are applied to optimize the size of communication ports,ensuring resources are allocated rationally and efficiently.In addition,the first-order derivative method was employed to calculate the optimal number of connection lines at the receiving end,opti-mizing the data transmission path and further diminishing the likelihood of congestion.Experimental outcomes dem-onstrate that the suggested approach effectively mitigates the risk of network communication congestion and maintains a low packet loss rate.关键词
数据中心/多对一通信模式/数据包优先级/缓冲区队列/连接线路Key words
data center/multi to one communication mode/packet priority/buffer queue/connecting lines分类
信息技术与安全科学引用本文复制引用
郁葱..高可扩展性数据中心多对一通信拥塞控制方法[J].苏州科技大学学报（自然科学版）,2026,43(1):72-80,9.基金项目
国家自然科学基金项目(51978239) （51978239）
