Mobile edge computing (MEC) can extend the computing capability of mobile users (MUs), by offloading computational tasks from MUs to proximal MEC servers. Specifically, offloaded tasks are firstly transmitted to access points (APs) via wireless channels, and then relayed to MEC servers through backhauls. Due to differentiated wireless channels and input data sizes, offloaded tasks from different MUs arrive at the AP and MEC server in an asynchronous manner. Considering such asynchrony, we design sequential execution model for backhaul transmission and task processing in the MEC server and present its merits in terms of maximum task completion time for different backhaul configurations: 1) With ideal backhaul, the optimal task scheduling order in the MEC server follows the first coming and first scheduling rule; 2) With limited backhaul, an effective two-stage task scheduling order on the backhaul and MEC server is determined by the Johnson’s rule. Further, we make suboptimal decisions on wireless resource provisioning and task offloading, with respect to the attained task scheduling order and derived necessary and sufficient conditions under which full local processing and full offloading are optimal. Finally, simulation results demonstrate the superiority of our devised sequential execution model and computation offloading mechanisms.

中文翻译：

---

多用户MEC系统中计算卸载的异步性

移动边缘计算 (MEC) 可以通过将计算任务从 MU 卸载到邻近的 MEC 服务器来扩展移动用户 (MU) 的计算能力。具体来说，卸载的任务首先通过无线信道传输到接入点（AP），然后通过回程中继到 MEC 服务器。由于无线信道和输入数据大小的差异，来自不同 MU 的卸载任务以异步方式到达 AP 和 MEC 服务器。考虑到这种不同步性，我们设计了 MEC 服务器中回程传输和任务处理的顺序执行模型，并从不同回程配置的最大任务完成时间方面展示了它的优点：1）在理想回程下，MEC 服务器中的最佳任务调度顺序遵循先到先调度规则；2) 回程有限，回程和MEC服务器上有效的两阶段任务调度顺序由约翰逊规则决定。此外，我们对无线资源提供和任务卸载做出次优决策，考虑到获得的任务调度顺序和导出的充分本地处理和完全卸载最佳的必要和充分条件。最后，仿真结果证明了我们设计的顺序执行模型和计算卸载机制的优越性。关于获得的任务调度顺序和导出的充分本地处理和充分卸载是最佳的充分必要条件。最后，仿真结果证明了我们设计的顺序执行模型和计算卸载机制的优越性。关于获得的任务调度顺序和导出的充分本地处理和充分卸载是最佳的充分必要条件。最后，仿真结果证明了我们设计的顺序执行模型和计算卸载机制的优越性。