2016 University ASD Call

This opportunity closed as of the deadline noted below.

Call for Proposals from NSF-Supported University Researchers

The ASD submission deadline is August 29, 2016

CISL is pleased to announce the availability of dedicated, large-scale computational resources from approximately January through March 2017 as part of its Accelerated Scientific Discovery (ASD) initiative. Approximately three to five NSF-supported, university-led projects from across the atmospheric sciences or the supporting computational sciences will be chosen. These projects will use NCAR’s newly installed Cheyenne system at the NCAR-Wyoming Supercomputing Center (NWSC).

Cheyenne Specs: The Cheyenne system will be a 5.34-petaflops SGI ICE™ XA cluster with 145,152 Intel Broadwell cores and 313 TB of memory. Its peak performance will be more than 3.5 times that of the Yellowstone supercomputer. See the Cheyenne home page for more information.

See this section below to compare Cheyenne and Yellowstone core-hours.

University ASD eligibility

University researchers in the atmospheric sciences and closely related sciences who have received and used a large allocation at NCAR or another HPC center are invited to apply. An active NSF award for the proposed research (or for which the proposed research is a natural extension) is required. Only one submission per research group will be accepted. Joint university/NCAR requests based on a university-led NSF award are eligible for ASD consideration.

ASD project requirements

To be successful, a proposed project must study a challenging and important problem in the atmospheric sciences or computational science in support of the atmospheric sciences. CISL is particularly interested in submissions for groundbreaking projects in line with the breakthrough computations described in the NWSC Community Science Objectives.

In addition to a challenging scientific problem, ASD projects must also:

  • Require 10 million core-hours or more for the set of computational experiments proposed. Very large requests are encouraged. To estimate the number of Cheyenne core-hours needed, multiply Yellowstone core-hours by 0.82. For example, using that factor, a 20 million core-hour project on Yellowstone would require 16.4 million core-hours on Cheyenne.

  • Utilize production-ready codes that can scale effectively and efficiently up to thousands of processors or efficiently use an ensemble of comparable size.

  • Have the staff time and experience to complete the computational experiments within 12 weeks.

Guidelines, submission, and schedule

All university submissions must follow the ASD university instructions and be submitted via the ASD university form by August 29, 2016. (NCAR-led projects must be submitted using a separate form since there are additional requirements.) University ASD projects will be reviewed and recommended by the CISL HPC Advisory Panel (CHAP).

ASD teams that are recommended for awards should be prepared to submit a preliminary benchmark from the Laramie test system by November 21, 2016, and a production benchmark that demonstrates efficient use of Cheyenne no later than December 30, 2016. These dates assume that first access to Cheyenne is on December 18, 2016. CISL HPC consultants will be available to provide tuning guidance and advice for ASD projects during the benchmark process.

Projects must begin production runs no later than January 3, 2017, and continue rapidly to completion by the end of March 2016. If a project is unable to meet these benchmark and production deadlines and guidelines, the ASD allocation will be reduced to 10% of the original allocation, and CISL will reallocate the resources.

CISL staff will contact recommended ASD projects to discuss data analysis and visualization options and opportunities before the project’s start on Cheyenne. ASD projects will be granted data analysis and visualization allocations through October 31, 2017.

ASD projects must provide a short preliminary progress report by February 13, 2017, and a final accomplishment report by November 30, 2017, of no more than five pages that describe the research results, the computational experiments performed, the code efficiency, and scientific papers submitted or published as a result of the ASD computations. Please read “Acknowledging NCAR/CISL” for how to properly acknowledge NSF and NCAR/CISL computational resource support in your publications.

Other opportunities

Researchers whose projects do not meet the size, scalability, readiness, and schedule criteria for ASD have other opportunities to apply for Cheyenne access. The university community will continue to have opportunities to submit large requests to the CHAP twice a year. The next deadline is September 12, 2016, for Cheyenne and other NCAR resources. See the University allocations page for details.

Questions

ASD request instructions and request form link are available here. Questions on the ASD allocation and selection process may be directed to alloc@ucar.edu or 303-497-2400. Questions about the new Cheyenne system and the user transition process may be directed to cislhelp@ucar.edu.