A Scalable Server-Side Solution for the Real-Time Handling of Road Safety Notifications

Abstract

Accidents are the main hurdles for using bicycles to change our transport habits. After many studies, there is no generally adopted solution. Two (non exclusive) approaches can be applied: one is based on the direct detection among vehicles, usually requiring additional hardware. The other consists of communicating through an external server which sends alerts to the concerned vehicles in real-time. In the latter case, if smartphones are used as the only instrumentation, the adoption of the system could be straightforward. In a previous work we validated the usage of conventional smartphones to create the client-side of a warning system. Instead, in the current work we address the server part. Such a server has to meet several requirements, such as being scalable (a matter not previously addressed), and able to meet real-time constraints. To achieve our purpose, we first provide the algorithms needed to ensure scalability. The system is composed by a dynamic pool of region servers, which controls a defined geographic area. Then, we implemented a functional prototype of such servers; its performance has been tested under realistic conditions to find the saturation point, after which real-time constraints are no longer guaranteed. Finally, the saturation point has been tested along with different traffic densities. Results show that the region server is able to track up to 15,000 simultaneous users, while the best we have found in published results are less than 1,400 users in simulated scenarios.