基于粒子群算法的边缘计算任务卸载研究

Research on Edge Computing Task Offloading Based on Particle Swarm Optimization Algorithm

随着物联网的高速发展,各类应用程序产生的数据大规模增长。传统的集中式网络在进行新型任务处理时经常存在因链路负载过重、时延过长等问题,导致任务执行时延过长甚至失败。移动边缘计算则可以通过将服务器资源下放至近用户端的同时采用相关任务调度策略来减少系统的时延,从而提高任务执行成功率。因此,边缘计算任务调度策略研究成为移动边缘计算领域的热点。本文针对移动边缘计算中的多用户任务卸载系统,以降低系统时延为目标,采用混合流水车间调度模型(Hybrid Flow-shop Scheduling Problem,HFSP)并结合粒子群优化算法(Particle Swarm Optimization,PSO),对多用户系统中多个独立任务的整体任务调度策略进行了研究。本文给出了在该模型下的系统时延表达式及其求解方法,获得了最优的任务执行甘特图,并得出与最低系统时延相匹配的最佳无线资源和计算资源分配方案,同时提高负载时系统时延增加不明显。

With the rapid development of Internet of Things, data generated by various applications explosively increases. In traditional centralized networks, task execution fails due to link overload and long delay. Mobile edge computing has been attracted more attention because it can reduce system delay and improve the success rate of task execution by transferring server resources to terminals near to user as well as using task offloading strategies. Aiming at the multi-user task unloading system in mobile edge computing, this paper adopted the HFSP and PSO algorithm to study the global task scheduling strategy of multiple independent tasks in multi-user system. The system delay was investigated and related equations for the delay were derived. Moreover, the Gantt chart of the optimal task execution was given. All the results showed that the minimum system delay agreed with the transmission rate of wireless resources and computing resources, and system delay did not increase significantly while the load increased.