Acquire, deploy, and maintain CI resources

High-performance and data-centric computing equipment, also called cyberinfrastructure (CI), is a foundational element of NCAR’s scientific research and publications. Consistent with NCAR’s mission, CISL deploys, operates, and maintains NCAR’s CI environment that is comprised of world-class petascale computing, data analysis, visualization, networking, data storage, and data archival systems used by about 2,500 researchers from both national and international institutions.

One of the core missions of CISL is the efficient design, operation, and maintenance of a robust, well-balanced and innovative computational and data environment in support of scientific productivity at research universities, NCAR, and other Earth System research organizations. CISL’s leadership in providing these discipline-focused computing and data services is a critical role for NCAR as a national center. The users of NCAR’s computational, data storage, and data archival services benefit from our data-centric paradigm for supercomputing that has been proven to be efficient, reliable, and user friendly for our existing and emerging computational and data-intensive applications and workflows.

The supercomputing and data storage CI at the NWSC was upgraded in early FY2017 with the addition of the 5.34-petaflops High Performance Computing (HPC) system named Cheyenne and a 40-petabyte increment to the GLADE shared parallel file system that more than doubled GLADE’s aggregate I/O bandwidth to 220 gigabytes per second. These resources began user production in January 2017. Cheyenne and its GLADE component, called GLADE-2, augmented the existing CI which included: Yellowstone, a 1.5-petaflops HPC system; GLADE, a 16.4-petabyte shared parallel file system; Caldera and Geyser, two separate data analysis and visualization (DAV) clusters featuring general purpose graphics processing units (GPGPUs); and NCAR’s HPSS-based data archive, a disk- and tape-based archival system currently holding more than 80 petabytes of data.

In addition to deploying new CI resources, CISL is also is actively evaluating future HPC and storage technologies with its HPC Futures Lab, an infrastructure to support the testing of experimental and early-release HPC hardware and software. In FY2017, CISL also launched an initiative to research the use of public and private cloud storage solutions and their feasibility to serve HPC environments by running a varied sample of user use cases on some of the established cloud service providers.

CISL’s goals and actions in this area directly support the broader goals of the NCAR Strategic Plan, CISL’s Strategic Plan, and the NSF’s Cyberinfrastructure Framework for 21st Century Science and Engineering (CIF21 Vision). The NCAR Strategic Plan states that “[HPC] is one of the foundational elements of NCAR’s scientific research and publications.” One of CISL’s key strategic imperatives is advancing Earth System science by expanding the productivity of researchers through HPC and data services. CISL therefore provides and refines a portfolio of advanced computing and data services specifically tailored for the atmospheric, geospace, and related sciences communities. CISL services evolve in response to changes in the underlying computational technologies and the scientific demands of the community, and are informed by the research and development activities performed under CISL Strategic Goal 2, namely, “to enhance the effective use of current and future computational systems by improving mathematical and computational methods for Earth System models and related observations.”

Cheyenne and its GLADE-2 resource were operated in full production status for the last nine months of FY2017, while Yellowstone and its associated complex of HPC and storage resources operated in full production status throughout FY2017. CISL will continue operating Yellowstone for the remainder of calendar year 2017, after which it will be decommissioned.

NCAR’s supercomputers are managed by CISL under the UCAR/NSF Cooperative Agreement and are supported by NSF Core funds.