MEC中面向动态环境的工作流D2D协同卸载方法  

Dynamic environment oriented D2D collaborative offloading for workflow applications in mobile edge computing

作　　者：钱程 赵淦森[1,2] 罗浩宇 QIAN Cheng;ZHAO Gansen;LUO Haoyu(School of Computer Science,South China Normal University,Guangzhou 510631,Guangdong,China;Key Lab on Cloud Security and Assessment Technology of Guangzhou,Guangzhou 510631,Guangdong,China)

机构地区：[1]华南师范大学计算机学院,广东广州510631 [2]广州市云计算安全与测评技术重点实验室,广东广州510631

出　　处：《山东大学学报（工学版）》2022年第4期45-53,共9页Journal of Shandong University（Engineering Science）

基　　金：广东省重点领域研发计划项目(2020B0101650001);国家自然科学基金项目(62002123);广东省基础与应用基础研究基金项目(2019A1515110212);国家重点研发计划项目(2018YFB1404402)。

摘　　要：为解决现实场景中终端设备的移动性与性能波动对工作流D2D(device-to-device)协同卸载带来的问题,提出一种面向动态环境的工作流D2D协同卸载方法,以尽可能小的卸载成本保证工作流在时间约束内得到及时响应。在工作流的卸载执行过程中感知其执行时间状态,并提出贪婪有序自适应搜索算法进行高效的工作流D2D卸载决策,根据环境变化及工作流执行时间状态在线调整卸载方案。仿真结果表明,该卸载方法在动态环境中具有有效性,并且卸载决策所需的计算开销很低(仅为粒子群优化算法的1.63%),具有较高的实时性。Aiming at the challenges of mobility and performance fluctuation of mobile devices in the real world,a dynamic environment oriented D2 D(device-to-device)collaborative offloading approach was proposed to ensure that the computational offloading requirements in edge devices could be responded with a minimum cost before the deadline.A method was proposed to perceive the temporal state of the workflow under the dynamic environment.The greedy ordered adaptive search procedure(GOASP)was proposed to make the decision-making more efficient.The offloading decision would be adjusted online with the above procedures.Experimental results showed that our proposed approach was effective in dynamic environments with low computational overheads(1.63%of the PSO-based approach)and high real-time performance.

关 键 词：移动边缘计算 工作流 D2D卸载 计算卸载 在线决策 

分 类 号：TP399[自动化与计算机技术—计算机应用技术]