NASA's High-End Computing Capability: It's Electra-fying!

Overview

NASA's High-End Computing Capability (HECC) Project provides the agency's primary supercomputing environment, focused on enabling science and engineering discovery across all technical mission areas—from Earth and space science to aerospace technology and space exploration.

This year, we began a process to significantly increase the power and cooling capability essential for future expansion of HECC resources by deploying Electra, NASA's first modular high-performance computing (HPC) system.

In addition, the HECC team increased the power of Pleiades, NASA's largest supercomputer, and expanded the storage capacity of the archive system to nearly a half-exabyte of data (compressed).

Project Details

In order to meet NASA's competing HPC demands—capacity, capability, and time-critical computing—the HECC team provides a stable, reliable environment that can be expanded to accommodate the agency's rapidly increasing needs for high-performance computing (HPC).

In 2016 we deployed the new 1.2-petaflop (PF) Electra system in a self-contained, energy-efficient module—a first step toward accomplishing the critical task of increasing the number of HPC systems HECC can support. This modular HPC technology promises to have a dramatic impact on the amount of water and electrical resources needed to cool our systems. With a power usage effectiveness (PUE) measurement close to 1.05, the system is designed to significantly reduce our impact on the environment as we continue to expand NASA's HPC capacity. Flexibility is achieved by acquiring and integrating additional modules over time, providing more compute resources as they are needed.

We also migrated the HECC archive storage library to the next-generation LTO-7 tape technology and increased Pleiades' theoretical peak performance to 7.25 petaflops—a 26% increase over the last year.

Results and Impact

Pleiades moved to its seventh generation of Intel Xeon architecture this year with the installation of 28 Broadwell racks. The system has now advanced through seven different processor technologies and three network interfaces while simultaneously delivering the equivalent of over 700,000,000 Westmere node-hours.

This year's highlights:

• Pleiades was ranked 15th in the June Top500 list of the world's most powerful supercomputers, and 9th on the High Performance Conjugate Gradient (HPCG) list, a benchmark that more closely models the types of work done by our user community.
• The Lou archive system was migrated to LTO-7 tape drives, enabling us to store up to a half-exabyte of compressed data in our library.
• The site was developed and the first module installed with the 1.2-PF Electra system, opening a gateway to further expansion.
• We worked with NASA Ames Research Center facilities teams and NASA headquarters to plot a growth path for HECC through 2022.

What's Next

HECC continues its quest to meet the growing demands of the NASA user community. The facility expansion that we began this year has strong potential for a huge upside in the project's ability to meet these demands. With advances in technologies and improvements in our approach to house and deliver computing cycles, we will continue to provide world-class computing to NASA's scientists and engineers across the agency.