Improved methodology for assessment of communication protocols for distributed road traffic simulation

Abstract

A computer simulation of road traffic is a commonly used tool, which can help to manage constantly intensifying road traffic. It can help to analyze behavior of existing road traffic networks or to predict the behavior of new designed road traffic structures. There are several existing simulators designed or adapted to run in a distributed computing environment in order to achieve a faster execution. In this environment, the inter-process communication ensured by a high-level communication protocol is one of the main bottlenecks limiting the speed of the entire computation. Various high-level communication protocols can have various efficiency, applicability, and scalability. This paper describes an improved methodology for testing and assessment of high-level communication protocols for micro-scale (or microscopic) distributed road traffic simulations. The methodology investigates the dependencies of the communication protocols’ performance on various features of the simulation and enables to easily calculate score for each of the tested protocols. Using the scores, the tested protocols can be directly compared. This can be useful when designing or improving a distributed road traffic simulation as the best protocol can be used in this simulation to improve its performance (e.g., its speedup or communication time). The improved version of the methodology is an evolution of its original version. It newly incorporates the assessment of the error introduced into the simulation by lossy communication protocols and reduces overall number of performed tests. The improved methodology was tested using a case study assessing several communication protocols of our own design.