物联网学报 ›› 2021, Vol. 5 ›› Issue (2): 78-86.doi: 10.11959/j.issn.2096-3750.2021.00199

所属专题： 边缘计算

• 专题：物联网边缘智能与雾计算技术 • 上一篇    下一篇

移动边缘计算任务切分与最优卸载算法设计

路静, 李晗琳, 高林   

  1. 哈尔滨工业大学（深圳）电子与信息工程学院，广东 深圳518055
  • 修回日期:2020-09-26 出版日期:2021-06-30 发布日期:2021-06-01
  • 作者简介:路静（1996- ），女，哈尔滨工业大学（深圳）电子与信息工程学院硕士生，主要研究方向为移动边缘计算中的任务卸载策略
    李晗琳（1996- ），女，哈尔滨工业大学（深圳）电子与信息工程学院硕士生，主要研究方向为移动边缘计算中的资源部署、任务卸载、经济模型分析等
    高林（1980- ），男，博士，哈尔滨工业大学（深圳）电子与信息工程学院副教授，主要研究方向为移动边缘计算、群智计算、群体智能、博弈论、强化学习等
  • 基金资助:
    国家自然科学基金资助项目(61972113)

Design of task dividing and offloading algorithm in mobile edge computing

Jing LU, Hanlin LI, Lin GAO   

  1. School of Electronic and Information Engineering, Harbin Institute of Technology(Shenzhen), Shenzhen 518055, China
  • Revised:2020-09-26 Online:2021-06-30 Published:2021-06-01
  • Supported by:
    The National Natural Science Foundation of China(61972113)

在线阅读

66

PDF下载

263

摘要：

移动边缘计算（MEC, mobile edge computing）作为将计算基础设施从远程云数据中心推向边缘设备的新架构模式，为满足物联网（IoT, Internet of things）应用时延敏感、计算密集等需求提供了新方案。针对可切分任务在多用户多 MEC 服务器系统中的任务卸载与调度问题进行研究，每个用户任务均可切分为多个相互关联的子任务，且子任务均可在本地执行或被卸载到某 MEC 服务器执行，系统通过对子任务的卸载和调度决策来提高网络性能。使用用户体验（QoE, quality of experience）和用户间公平性来表征网络性能，将优化问题建模为一个可切分任务卸载和调度（J-DTOS, joint dependent task offloading and scheduling）优化问题。该问题是一个NP-hard非线性混合整数规划问题，因此，所提方案进一步通过引入中间变量重新构造了原问题，并基于此提出了一个近似最优解。仿真结果表明，所提的卸载和调度策略能显著提高系统的性能。

关键词: 物联网, 移动边缘计算, 可切分任务卸载

Abstract:

Mobile edge computing (MEC) emerges as a new paradigm that pushes the computing infrastructure from the remote cloud data center to the edge equipments.It provides a new solution to meet the delay sensitive and computing intensive requirements of Internet of things (IoT).In this work, the problem of tasks offloading and scheduling in the multi-user and multi-server MEC system was considered.Specifically, each user had a task-dependent application and the tasks could be either executed locally or remotely according to the dependence.Thus, the network performance was improved by unloading and scheduling the sub tasks.Quality of experience (QoE) and fairness between users were used to characterize the network performance, and the optimization problem was modeled as a joint dependent task offloading and scheduling (J-DTOS) problem.The J-DTOS problem was a non-linear mixed integer programming, which was NP-hard in general.The original problem was reformulated by introducing intermediate variables and proposing a near-optimal solution.Simulation results show that the proposed offloading and scheduling design can significantly improve the performance of the system.

Key words: Internet of things, mobile edge computing, dependent task offloading

中图分类号: 

No Suggested Reading articles found!