DLMS: Acquisition of Instrumentation for a Digital Laboratory for Multi-Scale Science

建议没有。多尺度非线性物理是21世纪一些最紧迫、最激动人心的科学问题的新兴范式。这些包括预测湍流和多相流体流动，模拟生物气溶胶的大气输送，纳米工程材料和设备，以及探测物质的宇宙分布。长度尺度涉及的范围，从字面上看，从宇宙的大小到基本粒子的大小。高性能科学计算是解决这些具有挑战性问题不可或缺的工具。然而，模拟这些系统所产生的输出是如此之大，以至于在其来源长期存储数据一直是个问题。这导致了一种流行的研究风格，在这种风格中，非常有价值的科学数据只被使用一次，然后就被丢弃了。利用主要研究仪器基金，约翰霍普金斯大学的多尺度科学数字实验室（DLMS）将实现多尺度系统研究的新范式：使用集成数据库和集群作为“虚拟实验室”。该系统将允许对感兴趣的大规模系统进行快速内部模拟，并结合高达100tb数据的长期存储能力，大约每秒10gb的搜索速率，以及数据库本身快速灵活的计算分析工具。这将使我们能够接近我们完全无法达到的参数空间和物理域大小。在其他项目中，不可压缩流体湍流的模拟，前所未有的5500亿个点的网格大小，将在国家实验室和美国超级计算机中心进行，并在DLMS进行存储和分析。我们期望在跨多尺度系统的科学和工程领域的跨学科研究项目中，突破性的发展将得到促进。在DLMS工作的大约50名研究生、本科生和博士后将接受多尺度科学、数值模拟、并行计算、可视化、数据分析和数据库技术等领域的培训。我们将举办专业研讨会，向更多的科学家介绍为DLMS开发的工具。外展工作将把我们的研究成果和科学方法传播给当地的学校和社区。

Proposal No. CTS-0320907Principal Investigator: S. Chen, Johns Hopkins UniversityMulti-scale nonlinear physics is an emerging paradigm for some of the most pressing and exciting scientific problems of the 21st century. These include predicting turbulent and multiphase fluid flow, modeling atmospheric transport of bio-aerosols, nano-engineering materials and devices, and probing the cosmological distribution of matter. The length-scales involved range, literally, from the size of the universe down to the scale of elementary particles. High-performance scientific computing is an indispensable tool in tackling these challenging problems. However, the output generated by simulation of such systems is so large that long-term storage of the data at its source has been problematic. This has led to a prevalent style of research in which very valuable scientific data is used once and then discarded. Utilizing Major Research Instrumentation Funds, the Digital Laboratory for Multiscale Sciences (DLMS) at the Johns Hopkins University will realize a new paradigm in the study of multiscale systems: that of using an integrated database and cluster as a ``virtual laboratory''. This system will allow for fast in-house simulations of the large-scale systems of interest, combined with a long-term storage capability of up to 100 Terabytes of data, search rates of about 10 Gigabytes per second, and fast, flexible computational analysis tools within the database itself. This will allow us to approach parameter space and physical domain sizes that were entirely beyond our reach. Among other projects, a simulation of incompressible fluid turbulence, at an unprecedented grid size of 550 billion points, will be performed at national laboratories and US supercomputer centers and housed and analyzed at the DLMS. We expect that breakthrough developments will be facilitated in an interdisciplinary program of research in science and engineering across the spectrum of multiscale systems. About fifty graduate and undergraduate students and postdoctoral fellows working at the DLMS will receive training in the areas of multiscale sciences, numerical simulation, parallel computing, visualization, data analysis, and data-base technology. We will organize professional workshops introducing the tools developed for the DLMS to a wider audience of scientists. Outreach efforts will disseminate the results of our research and our scientific methods to the local schools and community.