1 Department of Computer Science , Brunel University London , Uxbridge , UK.
2 US Army Research Laboratory , Aberdeen Proving Ground , Aberdeen, MD , USA.
Philos Trans A Math Phys Eng Sci. 2019 Apr 8;377(2142):20180147. doi: 10.1098/rsta.2018.0147.
In the last few decades, multiscale modelling has emerged as one of the dominant modelling paradigms in many areas of science and engineering. Its rise to dominance is primarily driven by advancements in computing power and the need to model systems of increasing complexity. The multiscale modelling paradigm is now accompanied by a vibrant ecosystem of multiscale computing software (MCS) which promises to address many challenges in the development of multiscale applications. In this paper, we define the common steps in the multiscale application development process and investigate to what degree a set of 21 representative MCS tools enhance each development step. We observe several gaps in the features provided by MCS tools, especially for application deployment and the preparation and management of production runs. In addition, we find that many MCS tools are tailored to a particular multiscale computing pattern, even though they are otherwise application agnostic. We conclude that the gaps we identify are characteristic of a field that is still maturing and features that enhance the deployment and production steps of multiscale application development are desirable for the long-term success of MCS in its application fields. This article is part of the theme issue 'Multiscale modelling, simulation and computing: from the desktop to the exascale'.
在过去的几十年中,多尺度建模已成为许多科学和工程领域的主要建模范例之一。它的主导地位的上升主要是由于计算能力的提高和对越来越复杂系统建模的需求。多尺度建模范例现在伴随着一个充满活力的多尺度计算软件(MCS)生态系统,该系统有望解决多尺度应用程序开发中的许多挑战。在本文中,我们定义了多尺度应用程序开发过程中的常见步骤,并研究了一组 21 种有代表性的 MCS 工具在多大程度上增强了每个开发步骤。我们观察到 MCS 工具提供的功能存在几个差距,特别是在应用程序部署以及生产运行的准备和管理方面。此外,我们发现许多 MCS 工具都针对特定的多尺度计算模式进行了定制,尽管它们在其他方面是与应用程序无关的。我们得出的结论是,我们确定的差距是该领域仍在成熟的特征,并且增强多尺度应用程序开发的部署和生产步骤的功能是 MCS 在其应用领域取得长期成功的理想选择。本文是特刊“多尺度建模、模拟和计算:从桌面到 Exascale”的一部分。