不确定网络环境下的任务卸载和资源分配算法

Task offloading and resource allocation in an uncertain network

随着计算密集型和时延敏感型应用的大量涌现,在大数据和低时延计算需求的驱动下,移动边缘计算(mobile edge computing, MEC)在提升用户体验和降低能耗方面将发挥重要作用.然而,由于计算资源有限的MEC服务器无法快速响应海量突发的计算请求,计算任务在MEC服务器的排队等待时间是不可忽略且难以预测的.为了保证在排队等待时间不确定的网络环境中满足应用的计算时延需求,本文在计算时延约束条件下,以最小化系统能耗为目标,提出了一种基于随机模拟的任务卸载和资源分配两阶段随机规划算法SSSSP.仿真结果表明, SSSSP算法能够保证应用的计算时延需求,同时有效降低了系统能耗.

With the emergence of massive compute-intensive and delay-sensitive applications, and driven by big data and low delay computation requirements, mobile edge computing(MEC) will play an important role in improving user experience and reducing energy consumption. However, MEC servers have limited computation resources and cannot respond quickly to the large amounts of bursting computation requirements, so computation waiting time at MEC servers is unavoidable and unpredictable. To meet the computation delay requirements of the applications in an uncertain computation waiting time network, this paper formulates a system energy consumption minimization problem with computation delay constraints and proposes a stochastic simulation based two-stage stochastic programming(SSSSP) algorithm for task offloading and resource allocation. The simulation results show that the SSSSP algorithm can meet the computation delay requirements of the applications while effectively reducing the system energy consumption.