Parallel agent-based simulation with Repast for High Performance Computing

  title={Parallel agent-based simulation with Repast for High Performance Computing},
  author={Nicholson T. Collier and Michael J. North},
  pages={1215 - 1235}
In the last decade, agent-based modeling and simulation (ABMS) has been applied to a variety of domains, demonstrating the potential of this technique to advance science, engineering, and policy analysis. However, realizing the full potential of ABMS to find breakthrough research results requires far greater computing capability than is available through current ABMS tools. The Repast for High Performance Computing (Repast HPC) project addresses this need by developing a useful and useable next… 

Figures and Tables from this paper

Experiences in Developing a Distributed Agent-based Modeling Toolkit with Python
  • N. Collier, J. Ozik, E. Tatara
  • Computer Science
    2020 IEEE/ACM 9th Workshop on Python for High-Performance and Scientific Computing (PyHPC)
  • 2020
This work leverages the Numba, NumPy, and PyTorch packages and the Python C-API to create a scalable modeling system that can exploit the largest HPC resources and emerging computing architectures.
Repast HPC with optimistic time management
An existing optimistic time management mechanism, namely Time Warp, by Jefferson, is adapted for a distributed agent based simulation tool, namely Repast for High Performance Computing (Repast HPC), and a simple and self-adaptive technique for adjusting checkpoint intervals is incorporated.
Designing a benchmark for the performance evaluation of agent-based simulation applications on HPC
This paper provides an initial implementation of the benchmark for FLAME, FLAME GPU, Repast HPC and EcoLab, some of the most popular parallel ABMS platforms, and uses it for comparing the applications generated by these platforms.
Computing Platforms for Large-Scale Multi-Agent Simulations: The Niche for Heterogeneous Systems
The results confirm that the nature of dynamic interactions of autonomous agents among themselves, groups, and environments has most impact on performance of computing platforms and features heterogeneous systems as a promising platform for the even larger-scale MASs in the future.
An Agent-Based Simulation API for Speculative PDES Runtime Environments
This paper discusses a semantically-rich API allowing to implement Agent-Based Models in a simple and effective way, and presents an experimental assessment showing how this proposal allows to implement complicated interactions with a reduced complexity, while delivering a non-negligible performance increase.
An Agent Framework for High Performance Simulations over Multi-core Clusters
A novel approach to modeling and high-performance parallel simulation of scalable agent models based on actors and the Theatre agency is proposed and achieved through the parallel/distributed simulation of a large-scale multi-agent system.
Network Partitioning in Distributed Agent-Based Models
This dissertation demonstrates that distributed agent-based model simulations benefit from the use of graph partitioning algorithms that involve a local, neighborhoodbased perspective, and proposes two partitioning schemes that consider the bottom-up individual-centric nature of agent- based modeling.
Extreme-Scale Dynamic Exploration of a Distributed Agent-Based Model With the EMEWS Framework
The high-level programming model of the EMEWS framework is presented and it is demonstrated how it is used to integrate an active learning ME algorithm to dynamically and efficiently characterize the parameter space of a large and complex, distributed message passing interface agent-based infectious disease model.
High-Performance Computing Implementations of Agent-Based Economic Models for Realizing 1:1 Scale Simulations of Large Economies
We present a scalable high-performance computing implementation of an agent-based economic model using distributed + shared-memory hybrid parallelization paradigms, capable of simulating 1:1 scale
Toward the New Version of D-MASON: Efficiency, Effectiveness and Correctness in Parallel and Distributed Agent-Based Simulations
D-MASON framework is a distributed version of the MASON library for writing and running Agent-based simulations and provides a memory consistency modeling, where the previous state of the agent is made available (consistently) for all other agents and this is obtained by exploiting the Java Method Handler mechanism.


A Framework for Megascale Agent Based Model Simulations on Graphics Processing Units
This article investigates data-parallel computer architectures such as Graphics Processing Units (GPUs) to simulate large scale ABMs and believes that its system is the first ever completely GPU based agent simulation framework.
SWAGES-An Extendable Distributed Experimentation System for Large-Scale Agent-Based Alife Simulations
SWAGES allows for multi-language agent definitions, uses a general plug-in architecture for external physical and graphical engines to augment the integrated SimWorld simulation environment, and includes extensive data collection and analysis mechanisms, including filters and scripts for external statistics and visualization tools.
On Load Balancing for Distributed Multiagent Computing
A novel communication-based load-balancing algorithm that works by associating a credit value with each agent, in view of the special agent characteristics, is proposed, implemented, and evaluated.
µsik -- A Micro Kernel for Parallel/Distributed Simulation
Using the micro-kernel approach, a unified architecture for incorporating multiple types of simulation processes within a single parsimonious application programming interface (API) is developed.
Scaling time warp-based discrete event execution to 104 processors on a Blue Gene supercomputer
An overview is given of a unique PDES engine that has been designed to support Time Warp-style optimistic parallel execution as well as a more generalized mixed, optimistic-conservative synchronization, designed to run on massively parallel architectures with minimal overheads.
OpenMP parallelization of agent-based models
Parallel simulation of the Sharks World problem
Experimental results indicate that reasonable speedups may be achieved for a relatively small number of creatures; further, the speedups increased linearly as the number of Creatures in the system was increased.
A unified framework for conservative and optimistic distributed simulation
A parallel simulation protocol is proposed that allows different parts of a system to be simulated using different protocols, allowing these protocols to be switched dynamically, and a proof of correctness is presented.
Visual agent-based model development with repast simphony.
This paper presents a model of physical infrastructure network interdependency as an introductory tutorial and illustration of the visual modeling capabilities of Repast S, the new Repast Simphony platform.
A 'conservative' approach to parallelizing the Sharks World simulation
The authors describe how they parallelized a benchmark problem for parallel simulation, the Shark's World. The solution obtained is conservative, in the sense that no state information is saved and