In communication networks, edge computing is a method of distributing processing, storage, and bandwidth resources to the side of the network that is closest to the user. This could be achieved through small computing devices installed in proximity of user or the network routers. In this paper, we discuss the architecture and principles of mobile edge computing (MEC) and edge caching in large-scale wireless networks. Moreover, we also discuss the necessity, widespread use, and the future of MEC and caching technologies. Finally, we analyze five key issues when MEC and caching are used for large-scale wireless networks: (i) computation offloading, (ii) edge caching, (iii) multidimensional resource allocation, (iv) user association (including privacy protection), and (v) privacy protection. Furthermore, edge servers are typically equipped with limited resources and are unable to meet the service demands of all vehicular network users at the same time. Therefore, identifying locations for service offloading and providing low latency services to users, while working within the aforementioned constraints, continues to be a significant challenge. Using deep and deep intensive learning coordination, we propose an edge computing system model for 5G vehicular networks that includes an “end-to-edge cloud” coordination. We, then, suggest a distributed service offloading method, called DSOAC, based on the coordination. Based on data sets collected from real-world wireless communication networks, experimental findings reveal that the DSOAC method can decrease service offloading time by approximately 0.4 to 20.0% when compared to the four current service offloading techniques. We observed that this ranges from 4 to 20.4% of the typical user workloads. An approximate 4% of the average customer service delay was caused by the proposed offloading scheme.

在通信网络中，边缘计算是一种将处理、存储和带宽资源分配到离用户最近的网络一侧的方法。这可以通过安装在用户或网络路由器附近的小型计算设备来实现。在本文中，我们讨论了大规模无线网络中移动边缘计算 (MEC) 和边缘缓存的架构和原理。此外，我们还讨论了 MEC 和缓存技术的必要性、广泛使用和未来。最后，我们分析了 MEC 和缓存用于大规模无线网络时的五个关键问题：（i）计算卸载，（ii）边缘缓存，（iii）多维资源分配，（iv）用户关联（包括隐私保护）， (v) 隐私保护。此外，边缘服务器通常配备有限的资源，无法同时满足所有车载网络用户的服务需求。因此，在上述限制范围内工作的同时，确定服务卸载的位置并向用户提供低延迟服务仍然是一项重大挑战。使用深度和深度强化学习协调，我们提出了一种用于 5G 车辆网络的边缘计算系统模型，其中包括“端到端云”协调。然后，我们提出了一种基于协调的分布式服务卸载方法，称为 DSOAC。基于从现实世界无线通信网络收集的数据集，实验结果表明，DSOAC 方法可以将服务卸载时间减少大约 0.4 到 20。与当前的四种服务卸载技术相比，为 0%。我们观察到，这占典型用户工作负载的 4% 到 20.4%。大约 4% 的平均客户服务延迟是由提议的卸载计划造成的。