Edge-assisted content and computation-driven dynamic network selection for real-time services in the urban IoT

Abstract

Supporting city-wide exchange of information in Urban Internet of Things (IoT) systems using existing communication infrastructures is extremely challenging especially when traditional services operate in the same network resource. Additionally, the most advanced Urban IoT services focus on real-time data processing, which shifts the perspective and goal of the network when transporting data. In this paper, the notion of Quality of Computing (QoC) is introduced to capture the level of support the communication infrastructure provides to this family of computation applications. In this context, we propose a dynamic network selection mechanism based on Software Defined Networks (SDN) designed to provide QoC in Urban IoT scenarios where the heterogeneous network resources are shared. The proposed mechanism dynamically assigns portions of data from IoT streams over licensed and unlicensed bands to guarantee QoC while minimizing cost of operations and licensed band occupation. Instrumental to our technique is the recently proposed edge-computing architecture, where computational resources placed at the edge of wireless access networks enable the interconnection of network management to processing. We consider a real-time monitoring scenario, where sensors transmit a video stream which is processed to identify and classify objects. The supporting wireless infrastructure consists of WiFi that operates in unlicensed frequency bands and cellular communication technology, Long Term Evolution (LTE) operating in licensed bands. We demonstrate the performance by means of real-world experiments on a testbed with WiFi and LTE networks built with hostapd and OpenAirInterface.