Operate the NWSC efficiently and reliably

CISL will continue improving NWSC efficiency as measured by metrics such as its power utilization efficiency (PUE). Over the next five years, CISL expects to incrementally improve the NWSC’s exemplary PUE. Since 2013, the facility has performed exceptionally well and achieved many of its design goals for efficiency, maintainability, and reliability. In particular, NWSC’s design encompasses a number of energy-saving features representing the state of the art in efficiency and sustainability, and these features reduce operating costs significantly below those of traditional computing facilities. The design for NWSC assumed a 4-MW IT load and projected a design Power Utilization Effectiveness (PUE) of 1.08. With the facility at only 1.7 MW, it has frequently achieved a PUE of 1.11–1.15. The NWSC faces its next big challenge in 2016: the incoming Cheyenne supercomputer will increase the electrical load by almost 2 MW. With that additional load, it should be possible to meet or exceed the design PUE of 1.08. NWSC optimization and maintenance continues with lessons learned and minor modifications to reduce energy costs and improve efficiency.

NWSC computer room temperatures
The figure shows a three-dimensional temperature representation of the NWSC produced by a commercial Computational Fluid Dynamics application named TileFlow. A student assistant updated this work and provided a much more detailed model that we use to evaluate equipment installation scenarios and ensure that the NWSC operates as efficiently as possible.

When the petascale Yellowstone supercomputer was installed at the NWSC, NCAR fulfilled a top imperative of its previous strategic plan to meet the rapidly growing HPC needs of Earth System scientists. The effort reflected strong engagement by all stakeholders in a private-public partnership that included UWyo, the State of Wyoming, Cheyenne LEADS, the Wyoming Business Council, and NSF. NCAR's addition of the Cheyenne supercomputer extends our ability to meet researchers' growing needs into the future.

During summer FY2016 CISL employed a Mechanical Engineering intern from California Polytechnic State University to update and modernize the computational fluid dynamic models for all NCAR computing facilities. She performed a major update that produced the most detailed models to date of the NWSC. These models inform and allow the NWSC to optimize airflow performance and minimize operational costs.

Our ongoing collaboration with the University of Wyoming (UW) to evaluate and prototype automated continuous commissioning techniques is currently on hold for two reasons. SIParCS interns from last year worked with the software being proposed by UW and found a number of challenges with implementing the software at the scale of the NWSC. Secondly, the project funding anticipated by UW collaborators did not materialize.

The operational expenses for NWSC during FY2016 were met using NSF Core funds.