MRI: Acquisition of a Next Generation, Data-Centric Supercomputer

点击翻译按钮获取中文摘要

The project will support the acquisition of a data centric supercomputer at Rensselaer Polytechnic Institute (RPI). This instrument will lead to significant advancements in science and engineering problems currently being tackled at RPI's Center for Computational Innovations (CCI) for applications including: the definition of new designed materials, applying active flows control for energy savings and microbiological systems modeling for medical treatment planning. The research will also include the development of new extreme-scale simulation technologies, graph analysis algorithms and the construction of entirely new simulation workflows. Hundreds of researchers and students from over 20 universities, 5 DOE national laboratories, 3 major industrial research centers (Corning, GE and IBM), 50 faculty, 4 start-ups across 11 U.S. states will take advantage of this proposed cyberinstrument to continue making a deep impact on their research. Student participation has been key to CCI's current success and national interest is anticipated not only due to the instrument's ability to advance current research but also due to its potential as a prototype model for future exascale systems. Students engaged in projects supported by the instrument will become the next generation of compute and data intensive experts.The new instrument integrates IBM POWER9 CPUs with next generation NVIDIA Volta GPUs into a hardware accelerated unified memory system (e.g., cache coherent). Additionally, all compute nodes are augmented with non-volatile memory storage, and a subset of the nodes include FPGA acceleration. The system will be used by faculty, students and CCI collaborators to address current barriers caused by the need to interact with massive data volumes that are used in and produced by next generation simulation tools. The cyberinstrument and algorithmic developments to be carried out will enable a new level of understanding and enhance our ability to solve many key challenges including: the accurate diagnosis of breast cancer directly from large-scale image datasets; semantic integration of the abundance of heterogeneous, multimodal, and multiscale data to improve personal health; modeling plasmas in fusion reactors; modeling active flow control devices that will greatly increase the weather conditions under which wind turbines will produce electricity; and combined biological data and model integration on molecular, cellular, and organ levels to understand organism-level phenomena and gain predictive understanding in systems biology.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.