A CNN-Based Routing Scheme for Minimizing TCP Flow Completion Time in SD-DCNs

被引：0

作者：

Zhou, Yingjie ^{[1
]}

Xu, Mingchun ^{[1
]}

Chen, Yu ^{[1
]}

机构：

[1] Beijing Univ Posts Telecommun, Natl Engn Lab Mobile Network Technol, Beijing, Peoples R China

来源：

2023 26TH INTERNATIONAL SYMPOSIUM ON WIRELESS PERSONAL MULTIMEDIA COMMUNICATIONS, WPMC | 2023年

关键词：

Software-defined data center networks; flow completion time; CNN; routing;

D O I：

10.1109/WPMC59531.2023.10338843

中图分类号：

TP3 [计算技术、计算机技术];

学科分类号：

0812 ;

摘要：

Data centers play a vital role in supporting the increasing demand from the six-generation (6G) networks for cloud services, big data, artificial intelligence and other data-intensive tasks. Software-defined data center networks (SD-DCNs) represent a natural evolution of traditional data center architectures aimed at improving network utilization. In this paper, we consider routing optimization for minimizing the completion time of transmission control protocol (TCP) flows in SD-DCNs. Specifically, we propose a routing scheme based on convolutional neural networks (CNN). Under different traffic transmission scenarios in SD-DCNs, the simulation results show that the flow completion time (FCT) of our scheme is much shorter compared to the equal-cost multipath (ECMP) and the shortest path first (SPF) algorithm, particularly in scenarios involving high-demand applications.

引用

页码：224 / 229

页数：6