A scalable Edge Computing architecture enabling smart offloading for Location Based Services

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. (c) 2020 Elsevier B.V. All rights reserved.