The evolution of Location Based Services (LBS) is expected to play a significant role in the future smart city. The ever-increasing amount of data produced, along with the emergence of next-generation computationally intensive applications, requires new service delivery models. Such models should capitalize on the Edge Computing (EC) paradigm for supporting the data offloading process, by considering user’s contextual information in the offloading decision along with the infrastructure resource allocation operations, towards meeting the stringent performance specifications. In this article, a two-level Edge Computing architecture is proposed to offer computing resources for the remote execution of an LBS. At the Device layer, an initial offloading decision is performed taking into consideration the estimated position and quality of the wireless connection of each user. At the Edge layer, a resource profiling mechanism maps the incoming workload to EC computing resources under specific performance requirements of the LBS. Dealing with the dynamic workload, a scaling mechanism simultaneously takes the offloading decision and allocates only the necessary resources based on the resource profiles and the estimation of a workload prediction technique. For the evaluation of the proposed architecture, a smart touristic application scenario was realized on a real large-scale 5G testbed, following the principles of Network Function Virtualization (NFV) orchestration. The experimental results indicate the high accuracy of the localization technique, the success of the two-stage offloading decision and the scaling mechanism, while meeting the performance requirements of the LBS.

基于位置的服务（LBS）的演进有望在未来的智慧城市中发挥重要作用。产生的数据量不断增加，以及下一代计算密集型应用程序的出现，都需要新的服务交付模型。此类模型应利用边缘计算（EC）范式来支持数据卸载过程，方法是在卸载决策中考虑用户的上下文信息以及基础架构资源分配操作，以达到严格的性能规格。在本文中，提出了两级边缘计算体系结构，以提供用于LBS远程执行的计算资源。在设备层，考虑到每个用户的无线连接的估计位置和质量，执行初始卸载决策。在边缘层，资源配置机制根据LBS的特定性能要求将传入的工作负载映射到EC计算资源。在处理动态工作负载时，伸缩机制会同时做出卸载决策，并根据资源配置文件和工作负载预测技术的估计仅分配必要的资源。为了评估提出的架构，遵循网络功能虚拟化（NFV）编排的原理，在真正的大型5G测试平台上实现了智能旅游应用场景。实验结果表明定位技术具有很高的准确性，