节点文献

基于机器学习的网络能耗优化方法的设计与实现

Design and Implementation of Optimization Method for Network Energy Consumption Based on Machine Learning

【作者】 李钢

【导师】 陈山枝;

【作者基本信息】 北京邮电大学 , 计算机科学与技术, 2020, 硕士

【摘要】 随着互联网蓬勃发展以及网络用户数量与日俱增,网络能耗逐年增长,网络能耗优化已经成为国内外研究的热点问题之一。协同休眠技术通过流量迁移将网络流量集中到网络拓扑子集,并通过将空闲设备调整至休眠来有效降低网络能耗。但是,网络能耗优化过程也会增加网络传输时延,影响网络性能。因此,研究保障传输性能的协同休眠方法,对降低网络能耗和提高传输性能具有重要意义。论文总结了网络能耗优化技术的研究现状和典型方法,针对协作传输与网络能耗的组合优化问题,提出了一种基于机器学习的网络能耗优化方法,在兼顾传输性能的基础上降低网络能耗。论文将节能路由决策问题建模为多商品流问题,采用邻域搜索算法计算节能路由,实现网络流量迁移和空闲设备休眠,降低网络能耗。进一步,引入全连接神经网络,在输入维度扩增时保证节能决策过程的收敛性,实现可休眠设备序列的快速识别。最后,基于真实网络流量数据对神经网络进行训练,并分别从节能百分比、传输时延及计算速度等角度对所提网络能耗优化方法进行了仿真测试。仿真结果表明,论文所提网络能耗优化方法能够优化网络能耗,保证网络传输时延。

【Abstract】 With the rapid development of the Internet and the increasing number of network users,network energy consumption increases year by year.The optimization of network energy consumption has become one of the hot issues at home and abroad.Collaborative dormancy technology concentrates network traffic to a subset of the network topology through traffic migration and effectively reduces network energy consumption by adjusting idle devices to dormancy.However,this technology can also have an impact on network performance,such as increasing network transmission delay.Therefore,this paper studies a cooperative sleep method which can guarantee the transmission performance and it is of great significance to reduce network energy consumption and improve transmission performance.This paper summarizes the research status and typical methods of network energy consumption optimization technology.Aiming at the combinatorial optimization of cooperative transmission and network energy consumption,a new network energy consumption optimization method based on machine learning is proposed to reduce network energy consumption while considering the transmission performance.This paper models the energy saving routing decision problem as a multi-commodity flow problem and calculates the energy saving routing by using neighborhood search algorithm.And then,the proposed method realizes network traffic migration and idle device dormancy to reduce network energy consumption.Furthermore,a fully connected neural network is introduced to ensure the convergence of the energy saving decision-making process when the input dimension is amplified,so as to realize fast recognition of the sequence of dormant devices.Finally,based on the real network traffic data,the neural network is trained and the proposed method is tested from the aspects of energy saving percentage,transmission time delay and calculation speed.And the simulation results show that the proposed network energy consumption optimization method can optimize the network energy consumption and guarantee the network transmission delay.

节点文献中: 

本文链接的文献网络图示:

本文的引文网络