MRI-Consortium: Acquisition of a Supercomputer by the Front Range Computing Consortium

University Of Colorado At Boulder, Boulder CO

Investigators

Abstract

Proposal #: CNS 08-21794 PI(s): Tufo, Henry M. Julie, Keith; Loft, Richard D.; Mandel, Jan; Syvitski, James P. Institution: University of Colorado - Boulder Boulder, CO 80309-0572 Title: MRI/Acq.: Acq. of a Supercomputer by the Front Range Computing Consortium Project Proposed: This project, acquiring instrumentation to investigate and address the technical obstacles to practical petascale computing for CU-Denver, CU-Boulder, and NCAR, explores the interplay of multi-core, intra-socket and inter-node parallelism in a variety of applications and additionally studies extreme application scalability, (issues critical to the petascale era). The work enables the researchers from the Front Range Computing Consortium (FRCC) to make discoveries in multiple areas through the supercomputing instrument with peak performance (over 100 TeraFLOPs). Access to this critical state-of-the-art Track 3 computational instrumentation enables HPC-driven problems such as Earth-System Sciences (e.g., climatology, critical oceanography, meteorology, land surface dynamics), Biotechnology (e.g., genomics, structural biology, microbial ecology), renewable energy applications; Material Science (e.g., nano-structured composites), turbulence, turbulent magnetic dynamos, and laser-plasma acceleration, and Computational Science. The instrument proposed combines Sun?s highest density Constellation assemblies, Intel?s next generation Nehalem-EP microprocessors, a custom InfiniBand QDR-based mesh interconnect, NVIDIA accelerators, and Sun Thumper storage modules to produce a highly balanced I/O rich, and cost-effective computing system. This new instrument, deployed within the integrated framework of the FRCC, offers a broad spectrum of investigators the necessary computational ecosystem to advance their research and strengthen their collaborations with national laboratories. Because the HPC manifested by NSF cyber-infrastructure hierarchy calls for ?a significant number of systems with peak performance in the 50-500 teraflops range, deployed and supported on the local level by individual campuses and research organizations,? the instrument fits into the NSF cyber-infrastructure hierarchy as a track 3 system and will have a software environment that allows it to peer seamlessly with other NSF cyber-infrastructure investments. Exhibiting potential to benefit other Directorates and Offices, this work undoubtedly directly benefits GEO, MPS, BIO, ENG, OCI, and CISE. Broader Impacts: This project brings together a broad and diverse set of disciplines to focus on key research priorities. The Consortium also decreases the barrier for collaboration between researchers at institutions along the Front Range, thereby accelerating scientific discovery, increasing scientific productivity, and fostering vibrant HPC ecology. Moreover, the instrumentation facilitates focused education, training, outreach, and support. Courses offered through this initiative are a vital component of training the next generation of scientists and engineers in the areas of modeling and scientific computing, numerical analysis, applied analysis, and high performance computing. The instrumentation supports ten historically black colleges and universities through distance learning courses and leverages existing mentoring programs such as NCAR?s Summer Internship in Parallel Computational Science (SIParCS), UCAR?s Significant Opportunities in Atmospheric Research and Science program (SOARS), and the CU Boulder?s Summer Multicultural Access to Research Training (SMART) programs.