HECC: Meeting Today’s Needs, Planning for the Future

Overview

NASA’s challenging mission to explore space and understand the universe and the Earth within it requires the agency to innovatively apply and extend humankind’s most advanced capabilities, technologies, and knowledge. The High-End Computing Capability (HECC) Portfolio provides a suite of such powerful, leading-edge tools.

At its most fundamental level, the HECC Portfolio accelerates scientific discoveries and engineering achievements by providing resources and services that address NASA’s current computational and data analytics needs, and engages in research and development that will create the productive high-performance computing (HPC) environments of the future.

The HECC Portfolio is for the benefit of all NASA mission directorates to: advance the design of quiet, energy-efficient aircraft; safely launch the first woman and the next man to the Moon; gain new knowledge and understanding of Earth, solar system, and the universe; and develop transformative space technologies to enable future missions.

"To stay on the forefront of physics-based simulations and data analytic workload, HECC is preparing now for the paradigm changes that will be coming in the exascale era and beyond."

Project Details

HECC meets NASA’s current HPC requirements by providing world-class computing architectures and services that multiply the impacts of those assets. Each year, HECC determines the amount of compute time to be provided to its customers and tracks usage against the target’s allocations to the mission directorates and to NESC. Allocations can be used on HECC’s local or cloud-based assets.

Computer time is measured in wall-clock hours and is described in terms of Standard Billing Units (SBUs). One SBU is currently defined as the work accomplished during a wall-clock hour on one Intel “Broadwell” node. Since HECC systems comprise many model types, this standard allows work to be normalized for all users across all of the systems.

In addition to the compute resources, HECC users also have access to a suite of comprehensive, integrated services that include Supercomputing Systems, Application Optimization, Data Publication and Discovery, Visualization, Data Analytics, and Networking.

Results and Impact

This year, we continued our support of NASA’s science and engineering teams and began the effort of planning for the future. We conducted an agency-wide needs assessment for advanced computing, and tripled the size of our General-Purpose Graphics Processing Unit (GPGPU) cluster, providing a substantial platform for artificial intelligence and machine learning. The GPGPUs, along with ARM and vector processors, are being examined to assess their strengths and weaknesses for meeting the future needs of the agency. In addition, we expanded the Aitken supercomputer, growing our computational base with AMD technology. Because data sharing is also a growing area of interest for our user community, we now provides support for researchers to share their large datasets, hosted on HECC systems, with specified communities or with the public.

Finally, this year HECC completed the development of a cloud gateway for NASA users who want to run their jobs on advanced computing resources in the cloud—providing a new capability to the NASA community.

What's Next

While continuing to meet NASA’s current needs, HECC will partner with academia, industry, and other government agencies to prepare for a new generation of computers and ensure that future advanced computing technologies provide suitable platforms to support the agency’s engineering and research challenges. In addition to GPU-based systems—for which preparing applications is difficult and time consuming—the HPC industry will continue its focus on non-GPU machines. Many of these may use high-bandwidth memory (HBM) on the processor chips to improve performance or specialized processors for special domains such as artificial intelligence and machine learning.

For example, there are machines currently in design that will couple ARM processors with HBM. Much of NASA’s HPC workload, such as the OVERFLOW computational fluid dynamics code, are memory bandwidth bound and would benefit greatly from such an architecture. In this case, preparing applications is likely to be much more straightforward, but code and algorithm changes are likely still needed to take advantage of HBM and to reduce data traffic across the deep memory hierarchy.

At the same time, the agency’s need for data analytics is rapidly growing, driven by a substantial increase in the amount and sophistication of the data produced by next-generation satellites, large-scale modeling applications, and experimental environments such as wind tunnels—along with the increasing power of emerging data analytics tools.

To stay on the forefront of physics-based simulations and data analytic workload, HECC is preparing now for the paradigm changes that will be coming in the exascale era and beyond.

Related Links

• High-End Computing Capability