当前位置: X-MOL 学术Future Gener. Comput. Syst. › 论文详情
Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)
HPC-cloud native framework for concurrent simulation, analysis and visualization of CFD workflows
Future Generation Computer Systems ( IF 7.5 ) Pub Date : 2021-04-16 , DOI: 10.1016/j.future.2021.04.008
Carlos Peña-Monferrer , Robert Manson-Sawko , Vadim Elisseev

Analysis and rendering of high-resolution computational fluid dynamics (CFD) simulations often requires execution of multiple parallel data-processing pipelines. In this paper we present a hybrid cloud solution for efficient simulation and analysis of drop dispersions. The simulation component runs on a high-performance computing (HPC) cluster and cloud native framework performs the processing of CFD outputs. To facilitate some of the cloud features, we broke down the CFD data analysis pipeline into small microservice-like processes. In this way, a given resource can be set up and shared between different simulations, codes and locations to minimize idle times and dynamically adjust the capacity to the workload while producing results on the fly, independent of the simulation engine. A combination of HPC and cloud technologies allows us to create agile and highly scalable solutions for CFD purposes. We focus on HPC-cloud infrastructure; however, other arrangements such as cloud–cloud or desktop–cloud can be easily adapted depending on the needs. This enables a framework for centralizing services for collaborative use between different users as well as for automatically providing data from different sources to machine learning algorithms. We describe a proof of concept implementation of the proposed framework and provide detailed analysis of its performance applied to a real two-phase flow application.



中文翻译:

HPC云原生框架,用于并发仿真,分析和可视化CFD工作流

高分辨率计算流体动力学(CFD)模拟的分析和渲染通常需要执行多个并行数据处理管道。在本文中,我们提出了一种混合云解决方案,用于有效地仿真和分析液滴分散。仿真组件在高性能计算(HPC)集群上运行,云本机框架执行CFD输出的处理。为了促进某些云功能,我们将CFD数据分析管道分解为类似小型微服务的流程。通过这种方式,可以在不同的仿真,代码和位置之间建立并共享给定资源,以最大程度地减少空闲时间,并动态调整容量以适应工作负载,同时不依赖于仿真引擎而动态产生结果。HPC和云技术的结合使我们能够为CFD目的创建敏捷且高度可扩展的解决方案。我们专注于HPC云基础架构;但是,可以根据需要轻松调整其他安排,例如云(云)或桌面云(桌面云)。这为集中服务以在不同用户之间协作使用以及自动将来自不同来源的数据提供给机器学习算法的框架提供了框架。我们描述了所提出框架的概念验证实施,并提供了对其在实际两相流应用中的性能的详细分析。这为集中服务以在不同用户之间协作使用以及自动将来自不同来源的数据提供给机器学习算法的框架提供了框架。我们描述了所提出框架的概念验证实施,并提供了其在实际两相流应用中的性能详细分析。这为集中服务以在不同用户之间协作使用以及自动将来自不同来源的数据提供给机器学习算法的框架提供了框架。我们描述了所提出框架的概念验证实施,并提供了其在实际两相流应用中的性能详细分析。

更新日期:2021-04-27
down
wechat
bug