边缘环境下轨迹预测性感知的在线边缘服务分配  

Novel Predictive Approach to Trajectory-aware Online Edge Service Allocation in Edge Environment

作　　者：李晓波 陈鹏[2] 帅彬 夏云霓[1] 李建岐 LI Xiao-bo;CHEN Peng;SHUAI Bin;XIA Yun-ni;LI Jian-qi(Software Theory and Technology Chongqing Key Lab,Chongqing University,Chongqing 400044,China;Computer and Software Engineering,Xihua University,Chengdu 610039,China;Global Energy Interconnection Research Institute Co.Ltd.,Beijing 102209,China)

机构地区：[1]重庆大学软件与理论重庆重点实验室,重庆400044 [2]西华大学计算机与软件学院,成都610039 [3]全球能源互联网研究院有限公司,北京102209

出　　处：《计算机科学》2022年第11期277-283,共7页Computer Science

基　　金：国家电网信通院研究基金(52094020000U)。

摘　　要：移动通信技术的快速发展促使了移动边缘计算(Mobile Edge Computing,MEC)的出现。作为第五代(5G)无线网络的关键技术,MEC可利用无线接入网络就近提供电信用户所需服务和云端计算功能,从而创造出一个具备高性能、低延迟与高带宽的服务环境,加速网络中的各项内容、服务及应用。然而,如何实现MEC环境下有效且性能有保障的服务卸载和迁移仍然是一个巨大的挑战。针对这一问题,大多数现有的解决方案都倾向于将任务卸载视为一个离线决策过程,使用用户的瞬时位置作为模型输入。而文中考虑了一种预测轨迹感知的在线MEC任务卸载策略,即PreMig。该策略首先通过多项式滑动窗口模型对服务所属边缘用户的未来轨迹进行预测,然后计算用户在边缘服务器信号覆盖范围内的停留时间,最后以一种贪心策略进行边缘服务的分配。为了验证所设计的方法的有效性,基于真实MEC部署数据集和校园移动轨迹数据集开展了模拟实验,实验结果显示,所提策略在平均服务率和用户服务迁移次数两个关键性能指标上均优于传统策略。The rapid development of mobile communication technology promotes the emergence of mobile edge computing(MEC).As the key technology of the fifth generation(5 G)wireless network,MEC can use the wireless access network to provide the services and cloud computing functions required by telecom users nearby,so as to create a service environment with high performance,low delay and high bandwidth and accelerate various contents,services and applications in the network.However,it remains a great challenge to provide an effective and performance guaranteed strategies for services offloading and migration in the MEC environment.To solve this problem,most existing solutions tend to consider task offloading as an offline decision making process by employing transient positions of users as model inputs.In this paper instead,we consider a predictive-trajectory-aware online MEC task offloading strategy called PreMig.The strategy first predicts the future trajectory of edge users to whom the edge service belongs by a polynomial sliding window model,then calculates the dwell time of users within the signal coverage of the edge server,and finally performs the edge service assignment with a greedy strategy.To verify the effectiveness of the designed approach,we conduct simulation experiments based on real-world MEC deployment dataset and campus mobile trajectory dataset,and experimental results clearly demonstrate that the proposed strategy outperforms the traditional strategy in two key performance metrics,namely,the average service rate and the number of user service migrations.

关 键 词：边缘计算 移动性 移动轨迹预测 在线服务分配 服务迁移 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]