Premium
Validation of signal propagation modeling for highly scalable simulations
Author(s) -
Paciorek Mateusz,
Bujas Jakub,
Dworak Dawid,
Turek Wojciech,
Byrski Aleksander
Publication year - 2020
Publication title -
concurrency and computation: practice and experience
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.309
H-Index - 67
eISSN - 1532-0634
pISSN - 1532-0626
DOI - 10.1002/cpe.5718
Subject(s) - computer science , correctness , scalability , distributed computing , process (computing) , signal (programming language) , modeling and simulation , microscale chemistry , scale (ratio) , cluster analysis , similarity (geometry) , complex system , simulation modeling , simulation , algorithm , artificial intelligence , physics , mathematics education , mathematics , quantum mechanics , database , image (mathematics) , programming language , operating system , microeconomics , economics
Summary Efficient information flow in the complex, often microscale simulation systems such as the social, artificial life, or traffic ones poses a significant challenge. It is difficult to implement a highly scalable system due to algorithmic problems, which significantly hamper the efficiency, especially in the case of maintaining a synchronized state in a parallelized, distributed environment. Our previous work presented a desynchronized method of information distribution in a simulation environment, inspired by the propagation of smell, and proved this method to be highly scalable. In this paper, we enhance and validate this method to ensure it does not invalidate the conclusions drawn from the simulation, enabling the development of efficient, scalable simulation systems. The prototype of the method presented here leverages the actor model for parallelization and cluster sharding mechanisms for cluster management, providing a comprehensive solution for large‐scale simulations, following realistic rules known from the nature. In order to validate the method of signal propagation modeling, three simulation models are created and tested. The validation is based on statistical analysis of metrics collected during the simulation execution. Statistical similarity of the results obtained from the distributed and nondistributed executions indicates that the distribution process does not impact the correctness of the simulation.