Breakthrough Molecular Dynamics Research via an Anton2 Supercomputer

 描述(由申请人提供)：该项目代表了匹兹堡超级计算中心(PSC)和D.E.Shaw Research(DESRES)之间长期、持续和富有成效的合作伙伴关系的延续。该项目的首要目标是为生物医学研究社区提供访问由DESRES设计和开发的真正独特的超级计算资源ANTON的机会。Anton的性能比目前的通用超级计算机系统高出大约两个数量级，使生物医学科学家能够以前所未有的规模进行常规的长时间尺度分子动力学(MD)模拟。最近，下一代Anton 2架构在2014年超级计算大会上公开亮相，并与Anton 1一样，获得了享有盛誉的ACM Gordon Bell奖。自2010年10月以来，DESRES已经在PSC免费提供了一台512节点的Anton 1机器，用于非营利性用途。这台机器可以以高于10微米S/天的速率模拟含有数万个原子的系统--大约比在最快的通用机器上运行的最先进的并行MD模拟包所实现的速度大两个数量级。到目前为止，120个独特的私人投资机构已经对这一特殊资源进行了229个研究项目。ANTON上可能的研究范围与MD研究界本身一样广泛，包括蛋白质折叠、离子通道选择性和门控、蛋白质-膜信号以及药物与蛋白质的相互作用等。这种广度反映在迄今为止已经在机器上完成的一系列项目中。科学界在PSC使用ANTON已经产生了大量重要的研究成果和影响很大的出版物，我们认为如果不这样做是不可能的。在这里，我们建议继续向生物医学研究社区提供Anton机器。由于PSC的Anton是DESRES实验室之外的唯一一台机器，该项目为科学家提供了一个独特的机会来访问这一资源，以进行目前可能进行的最长时间尺度的MD模拟。重要的是，在获奖的第二年，我们计划用128节点的Anton 2计算机升级Anton 1，这将显著提高计算能力。与目前的ANTON一样，将由PSC托管的ANTON 2机器将由DESRES免费提供，因此对于生物医学研究社区来说，这是一项数百万美元的资源，PSC只要求提供资金来支付运营费用、用户支持、文件存储和基本基础设施成本。这次升级将使模拟速度提高约4倍，重要的是，科学家将能够模拟高达70万个原子的系统大小，而安东1号的极限是15万个原子。因此，Anton 2将为全国各地的生物分子科学家提供无与伦比的模拟能力，使生物假说的研究成为可能。

 DESCRIPTION (provided by applicant): This project represents the continuation of a longstanding, ongoing and very productive partnership between the Pittsburgh Supercomputing Center (PSC) and D. E. Shaw Research (DESRES). The project's overarching goal is to provide the biomedical research community access to a truly unique supercomputing resource, Anton, designed and developed by DESRES. Anton's performance exceeds that of present general-purpose supercomputer systems by approximately two orders of magnitude, and allows biomedical scientists to conduct routine long timescale molecular dynamics (MD) simulations at an unprecedented scale. Recently, the next-generation Anton 2 architecture was publicly unveiled at the 2014 Supercomputing Conference and, like Anton 1, was awarded the prestigious ACM Gordon Bell Prize. Since October 2010, DESRES has made a 512-node Anton 1 machine available without cost at the PSC for not-for-profit use. This machine can simulate systems with tens of thousands of atoms at rates above 10 µs/day - roughly two orders of magnitude greater than typically achieved by state-of-the-art parallel MD simulation packages running on the fastest general-purpose machines. So far 229 research projects by 120 unique PIs have been conducted on this exceptional resource. The range of investigation possible on Anton is as broad as that of the MD research community itself, and includes protein folding, ion channel selectivity and gating, protein-membrane signaling, and drug interactions with proteins, to name but a few. This breadth is reflected in the range of projects that to date have been accomplished on the machine. Use of the Anton at PSC by the scientific community has already resulted in a significant number of important research findings and high impact publications that we believe would not have been possible otherwise. Here, we propose to continue to make an Anton machine available to the biomedical research community. Since the Anton at PSC is the only machine outside of the DESRES lab, this project represents a unique opportunity for scientists to access this resource for conducting the longest timescale MD simulations currently possible. Importantly, in Year 2 of the award we plan to upgrade Anton 1 with a 128-node Anton 2 machine that will provide a significant increase in computational capabilities. As with the current Anton, the Anton 2 machine to be hosted by PSC will be generously made available by DESRES without cost and thus represents a multi-million dollar resource for the biomedical research community for which the PSC is only requesting funds to cover operating expenses, user support, file storage and basic infrastructure costs. This upgrade will provide roughly a 4-fold increase in simulation speed and, importantly, will allow scientists to simulate system sizes of up to 700,000 atoms, compared to the Anton 1 limit of 150,000 atoms. Thus, Anton 2 will provide biomolecular scientists across the nation with unparalleled simulation capabilities, enabling the study of biological hypotheses not otherwise possible.