Mid Sweden University

The Smart City (SC) paradigm is based on the integration of Information and Communication Technology (ICT) into the urban asset, for the optimal management of the energies and resources. The Internet of Thing (IoT) technology seems the proper solution to achieve this target, thanks to its capability to abstract the object in the real world. The deployment of IoT devices at different level in urban infrastructures is causing the presence of thousands of intelligent devices, large part of them with unused computational capabilities. Such devices could be integrated in a cluster in order to share the unused resources with other devices with limited computational resources. The use of a cluster of IoT Sensors has several benefits, including, but not limited to: high availability, sharing of computational resources, reduced response time with the respect of centralized cloud computing solution. The main bottleneck of this approach is represented by the communication infrastructure, typically based on wireless connection and, thus with a limited available bandwidth. The aim of the work related to this paper is to analyze the impact the communication infrastructure has on computational performance of a cluster of IoT sensors. An experimental set-up for the characterization of the performance of a cluster of low-cost off-the-shelf devices has been described. The experimental validation highlighted as the network infrastructure is loaded only during the data transfer and the maximum network load, with a cluster of ten IoT nodes is approximately 2 Mb/s with the considered benchmark. © 2019 IEEE.