Collaborative Research: PRObE - The NSF Parallel Reconfigurable Observational Environment for Data Intensive Super-Computing and High End Computing

并行可重构观测环境(Probe)支持对未来超级计算体系结构的研究。Probe为底层系统研究社区提供了必要且独特的基础设施，以研究可伸缩性(每个节点的线程、节点和处理器数量)以及未来超级计算体系结构面临的其他挑战。这一基础设施填补了国家能力方面众所周知的重大长期研究设施缺口，它为系统和低水平研究创建了一个非常大的共享资源，允许故障注入、电源控制、控制直到硬件的整个资源，并提供对资源的直接动手控制。探测器的独特能力为研究和动手实验开辟了新的机会，这在学术界以前是不可行的。使用Probe，低级系统研究界正在改进对高端计算的系统级挑战的理解，并创造有效的技术来解决这些挑战。Probe为学术界创造了新的研究机会，具有对未来超级计算架构产生变革性影响的巨大潜力。Probe使用Probe试验台为来访的HEC系统学生提供实习计划和每年一次的暑期学校实践培训。拟议中的努力包括特别关注来自代表性不足群体的学生，包括新墨西哥州当地人口。拟议的基础设施有可能促进基础知识进步的发展，这是美国保持其在计算系统、数据密集型云计算和数据中心自动化方面的领先地位所需的。

A Parallel Reconfigurable Observational Environment (PRObE) supports research on future supercomputing architectures. PRObE provides a needed and unique infrastructure to the low-level systems research community to study scaling (number of threads, nodes, and processors per node) and other challenges facing future supercomputing architectures. This infrastructure fills a well known significant long standing research facilities gap in the nations capabilities by creating a very large shared resource for systems and low-level research that allows fault injection, power control, controlling the entire resource down to the hardware, and providing direct hands on control of the resource. The unique capabilities of PRObE open up new opportunities for research and hands-on experimentation that is was previously not feasible in the academic community. Using PRObE the low level systems research community is developing an improved understanding of the systems-level challenges for high-end computing, and the creating effective techniques to address them. PRObE creates new research opportunities for the academic community that has the significant potential to have a transformative impact on future supercomputing architectures. PRObE provides an internship program and an annual summer school for hands-on training for visiting students on HEC systems using the Probe testbed. The proposed effort includes special attention to students from underrepresented groups, incorporating populations local to the New Mexico location.The proposed infrastructure has the potential to facilitate the development of advances in basic knowledge required for the US to maintain its leadership positions in computing systems, data-intensive cloud computing, and data center automation.