Design and Evaluation of a Distributed Architecture for GeoSparkSim

Abstract

The human being throughout history has developed ways to optimize his time and resources. Nowadays, mobility is one of the biggest problems. Traffic simulation tools are frequently used to study mobility problems. GeoSparkSim is a vehicular traffic simulation tool that extends Apache Spark to generate datasets of road network traffic significantly with the simulation of microscopic traffic that includes traffic lights, change of track, and the pursuit of vehicles. This work presents an architecture to implement and evaluate the GeoSparkSim simulation tool. The architecture was developed on different containers in Docker, and two stages of the study were created for its evaluation. Each stage contains 26 simulations in different geographic zones from the Quito city. Results show the potentials of GeoSparkSim in order to study traffic mobility and compare its implementation on different Cloud services such as Google, Amazon, and Azure. In order to have an acceptable simulation, we found a machine that has at least 146 GB in storage and 8 GB in RAM is recommended.