arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2602.15317cs.MA
收藏

提升NetLogo计算效率:在AWS和云基础设施上运行大规模基于代理模型的最佳实践

Enhancing Computational Efficiency in NetLogo: Best Practices for Running Large-Scale Agent-Based Models on AWS and Cloud Infrastructures

RTI International(RTI国际)

Michael A. Duprey, Georgiy V. Bobashev

AI总结:

本文介绍了在AWS和云基础设施上运行大规模基于代理模型的最佳实践,通过优化NetLogo的内存管理、Java选项和实例选择,实现了计算成本降低和性能提升。

摘要:

随着基于代理模型(ABMs)的复杂性和规模的增加,需要高效的计算策略来管理日益增长的处理能力和内存需求。本文提供了一本全面指南,指导如何优化NetLogo,该平台广泛用于ABMs,以在亚马逊网络服务(AWS)和其他云基础设施上运行大规模模型。它涵盖了内存管理、Java选项、BehaviorSpace执行和AWS实例选择的最佳实践。通过实施这些优化并选择合适的AWS实例,我们实现了32%的计算成本降低和性能一致性的提高。通过使用狼-羊捕食模型在不同AWS实例上比较NetLogo模拟,我们展示了通过这些优化所能实现的性能提升。

英文摘要:

The rising complexity and scale of agent-based models (ABMs) necessitate efficient computational strategies to manage the increasing demand for processing power and memory. This manuscript provides a comprehensive guide to optimizing NetLogo, a widely used platform for ABMs, for running large-scale models on Amazon Web Services (AWS) and other cloud infrastructures. It covers best practices in memory management, Java options, BehaviorSpace execution, and AWS instance selection. By implementing these optimizations and selecting appropriate AWS instances, we achieved a 32\% reduction in computational costs and improved performance consistency. Through a comparative analysis of NetLogo simulations on different AWS instances using the wolf-sheep predation model, we demonstrate the performance gains achievable through these optimizations.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑