Performance-Driven Analysis for an Adaptive Car-Navigation Service on HPC Systems

Car-navigation system was recently established as an imperative utility for modern navigation on road networks. The rising wave of self-driving cars along with an increasing demand for real-time traffic data is expected to generate massive growth of routing requests and processing time on large graphs representing the urban networks. Therefore, larger and more powerful computing infrastructures are required. In the context of smart cities, new, dynamic solutions are needed to deliver high-quality car-navigation services, powered by municipal traffic-monitoring data, capable of handling such a vast expected demand with reasonable employment of financial resources. In this work, we introduce an adaptive car-navigation system and its performance model used to tune the size of the computing infrastructure as a function of the characteristics of the environment considered. The model has been validated for a smart city environment using the data collected on the Milan urban area.