基于SDN的车联网多MEC动态负载均衡算法  

Software-Defined Networking-Based Dynamic Load Balancing Algorithm for Vehicular Networks with Multiple Mobile Edge Computing

作　　者：吴涛[1] 欧阳 周启钊 陈曦[2] WU Tao;OU Yang;ZHOU Qizhao;CHEN Xi(Computer Science Academy,Chengdu University of Information Science and Technology,Chengdu 610225,China;Computer Science Academy,Southwest Minzu University,Chengdu 610041,China)

机构地区：[1]成都信息工程大学计算机学院,四川成都610225 [2]西南民族大学计算机学院,四川成都610041

出　　处：《软件导刊》2024年第11期116-124,共9页Software Guide

基　　金：四川省科技计划项目(2021YFQ0056,23ZDYF0171,23RKX0645);成都信息工程大学引进人才科研启动项目(KYTZ202269)。

摘　　要：车载自组织网络(VANET)承载的数据规模呈现爆炸性增长趋势。针对车联网中在线卸载场景下,多边缘服务器(MEC)负载不均衡导致车辆卸载成功率严重下降问题,提出一种基于软件定义网络(SDN)的车联网多MEC动态负载均衡算法DFPC。该算法结合排队论中先到先服务和有优先权的服务两种方式,SDN控制器通过一定的等待时延定时收集当前批任务,利用改进的K-means聚类算法快速对多维任务分类,优先入队紧急度相对高的任务;再利用SDN控制器定时收集的MEC上下文信息,实现卸载任务在多个MEC之间分配的动态反馈调节,解决了多MEC之间动态负载不均衡问题,充分利用MEC的计算资源,最终提升了整体车辆卸载成功率。为了验证DFPC算法在真实动态场景下的有效性,设计一种多MEC接入的在线卸载框架MOLF,通过低成本硬件部署模式完成在线卸载场景下负载均衡性能测试。实验结果表明,相比基准方案,DFPC算法平均卸载成功率提升了28%,平均负载方差降低了73%。The data volume carried by vehicle ad-hoc networks(VANETs)exhibits an explosive growth trend,targeting the issue of a signifi⁃cant decline in offloading success rate due to uneven loads on multi-access edge computing(MEC)servers,this study proposes a software-de⁃fined networking(SDN)-based vehicular network multi-MEC dynamic load balancing algorithm DFPC.This algorithm combines the two meth⁃ods of first-come-first-serve and priority service in queuing theory,where the SDN controller periodically collects the current batch of tasks after a certain waiting delay.It uses an improved k-means clustering algorithm to rapidly categorize multidimensional tasks,prioritizing tasks with higher relative urgency for queuing.And using the MEC context information collected by the SDN controller at regular intervals,it imple⁃ments dynamic feedback adjustment for task distribution among multiple MECs,resolving the issue of dynamic load imbalance among multiple MECs.This approach optimally utilizes MECs computational resources,ultimately enhancing the overall success rate of offloading for vehicles.To validate the effectiveness of the DFPC algorithm in real dynamic scenarios,a multi-MEC online load balancing framework,named MOLF,is designed.Performance testing for load balancing in online offloading scenarios is conducted through a low-cost hardware deployment model.Experimental results indicate that,compared to the baseline solution,the DFPC algorithm increased the average offloading success rate by 28%and reduced the average load variance by 73%.

关 键 词：车载自组织网络 移动边缘计算 负载均衡 在线卸载 软件定义网络 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程] TN92[自动化与计算机技术—控制科学与工程]