MRI: Acquisition of High Performance Hybrid Computing Cluster to Advance Cyber-Enabled Science and Education at Penn State

在这个数据驱动的新时代，计算机现在正在成为突破性发现的驱动力，并正在改变科学和教育。宾夕法尼亚州立大学的天文、材料和物理网络实验室(CyberLAMP)将组建一个尖端的超级计算机集群，其中包括传统的中央处理器(CPU)和最新的硬件加速器，如图形处理器(GPU)，以推动网络科学的跨学科研究和教育。天文学家和物理学家将使用这台高性能的混合计算机来分析来自革命性的测量和实验的数据，并执行最先进的模拟来揭开我们宇宙的起源。材料科学家将进行真实、原子尺度的模拟，以指导下一代复杂材料的设计和开发。计算机科学家将分析这些科学应用，为未来计算机体系结构的设计提供信息。数据分析和模拟方面的这些进展将使CyberLAMP成员能够对国家战略计划优先考虑的主题有新的了解，例如国家研究委员会2010年的天文和天体物理十年调查，以寻找宜居行星，了解宇宙的基础物理，以及白宫的材料基因组计划，以加快新材料的开发。此外，CyberLAMP团队将利用这一集群来加强广泛的外联计划，包括：为宾夕法尼亚州立大学(包括其英联邦校区)的众多学生提供计算机教育；为研究人员和高中教师举办暑期讲习班；以及与业界建立伙伴关系，以推进材料研究和未来硬件-软件系统的联合设计。通过加快探索性数据分析和模拟，并在开发和原型算法方面促进跨学科合作，新的混合集群将使CyberLAMP团队能够在一些关键研究领域实现变革性突破。这包括用于世界领先的宇宙学和系外行星调查的最先进的天体统计学和天体信息学的数据分析，以及直接解决暗物质和暗能量的性质以及行星系统形成的复杂模拟；用于探索基础物理的最雄心勃勃的天体物理实验的重建算法的速度提高了数量级，这将扩大中微子、重力波和多信使发射器的宇宙源的发现空间，并提高对中微子质量等级的敏感性；在纳秒尺度的完全反应分子动力学模拟方面取得显著进展，用于开发下一代复杂材料；以及设计混合系统中的下一代硬件加速器的新颖见解、高度并行的算法和软件接口，这些可能会彻底改变数据密集型应用程序使用硬件加速器的方式。

Computers are now becoming the driving forces for ground-breaking discoveries and are transforming the science and education in this new data-driven era. The Cyber-Laboratory for Astronomy, Materials and Physics (CyberLAMP) at Penn State will put together a cutting-edge supercomputer cluster that includes both traditional central processing units (CPUs) and the latest hardware accelerators, such as graphics processing units (GPUs), to advance interdisciplinary research and education in cyberscience. Astronomers and physicists will use this high-performance hybrid computer to analyze data from revolutionary surveys and experiments and to perform state-of-the-art simulations to unravel of the origin of our Universe. Material scientists will run realistic, atomistic-scale, simulations to guide the design and development of next-generation complex materials. Computer scientists will analyze these science applications to inform the design of future computer architectures. These advances in both data analysis and simulations will enable the CyberLAMP members to shed new light on topics prioritized by national strategic plans, such as National Research Council's 2010 Decadal Survey for astronomy and astrophysics to search for habitable planets and to understand the fundamental physics of the cosmos and the White House's Materials Genome Initiative to expedite development of new materials. Furthermore, the CyberLAMP team will employ this cluster to enhance a wide range of outreach programs including: computational education to numerous students at The Pennsylvania State University, including its Commonwealth campuses; summer workshops for researchers and high-school teachers; and partnerships with industry to advance materials research and the co-design of future hardware-software systems. By expediting exploratory data analysis and simulations and catalyzing cross-disciplinary collaboration in developing and prototyping algorithms, the new hybrid cluster will enable the CyberLAMP team to deliver transformative breakthroughs in a number of key research areas. This includes state-of-the-art astrostatistics and astroinformatics for data analysis for world-leading surveys in cosmology and exoplanets, as well as sophisticated simulations to directly address the nature of dark matter and dark energy, and the formation of planetary systems; an order-of-magnitude increase of speed for reconstruction algorithms for the most ambitious astrophysical experiments probing fundamental physics, which will enlarge the discovery space for cosmic sources of neutrinos, gravity waves, and multi-messenger emitters, as well as heighten sensitivity to the neutrino mass hierarchy; dramatic advances in nanosecond-scale fully reactive molecular dynamics simulations for the development of next-generation complex materials; and novel insights for designing the next-generation of hardware accelerators in hybrid systems, highly parallel algorithms and software interfaces which could revolutionize the way hardware accelerators are used by data-intensive applications.