The Need for Increased Exascale Dollars

High Performance Computing (HPC) is the backbone of research and development in various scientific disciplines, astrophysics, engineering and biology, to name a few. In fact, the enhanced speed and complex computing of HPC is a foundation in scientific innovation, international competition and economic prosperity successes, and helps address national priorities such as climate change, national security and public health, including the ongoing COVID-19 pandemic.

Exascale computing is the next level of HPC, with capabilities to perform quintillion operations a second. Exascale computing systems are poised to increase memory, storage and compute power for critical and complex computing problems.

The Department of Energy is hard at work to deploy the first exascale systems in the world: Frontier, Aurora and El Capitan.

System Name	Location	Build	Expected Delivery	Expected Capabilities
Frontier	Oak Ridge National Laboratory	Cray architecture and AMD processor	2021	Greater than 1.5 exaflops of performance
Aurora	Argonne National Laboratory	Cray architecture and Intel processor	2022	50 times more computational power than leading HPC systems
El Capitan	Lawrence Livermore National Laboratory	Cray architecture and AMD processor	2023	2 exaflops of performance

A recent article cites Morgan McCorkle, Media Relations Manager at Oak Ridge, confirming that Frontier is on track for delivery by the end of 2021 with user operations scheduled in 2022. On day one, two dozen scientific and technical disciplines will run on Frontier, according to Dr. Georgia Tourassi, Director of the National Center Computational Sciences at Oak Ridge. Applications will provide research in areas including quantum, chemistry, additive manufacturing, fuels, fusion, fission, and more.

Just how much will be invested in U.S. exascale efforts? A look at the Department of Energy’s FY 2022 budget requests gives provides some fair insight:

Observations:

• Budget numbers reflect the partnership of the Office of Science (SC) and National Nuclear Security Administration (NNSA) under DOE’s Exascale Computing Initiative (ECI). ECI is comprised of three components
  • The Exascale Computing Project (ECP): the research, development, and deployment of the exascale applications and software ecosystem
  • Additional investments by SC and NNSA for mission-specific work
  • Deployment of the three exascale systems
• $150M requested in FY 2022 for the Argonne Leadership Computing Facility
• $125M requested in FY 2022 for the Oak Ridge Leadership Computing Facility, a $5M increase over FY 2021
• ECP FY 2022 funding breakdown:
  • $15M for Applications
  • $15M for Software
  • No funding requested for hardware, a decrease of $40M from FY 2020

As the above budget numbers reveal, investment in exascale computing is sliding. Though decreases are somewhat attributed to the imminent deployment of DOE’s first exascale systems, significant new exascale funding at DOE does not seem apparent.

However, Dr. Tourassi testified before the House Subcommittee on Science, Space and Technology in May 2021 making a case to raise spending in exasacle computing in the U.S.

She explained that researchers were already submitting worthy computational proposals three to five more times the hours than the current HPC program can award. Those that harness advanced computation, she argued, will make the fastest discoveries and innovations over global competitors.

The Oak Ridge director provides three key steps to advancing U.S. computing efforts:

1. Integrate Advanced Computing Technologies: exascale systems with AI, advanced analytics and modeling and simulation will identify data patterns beyond traditional approaches
2. Secure National Data Infrastructure: improvement in national R&D capabilities are prudent to compute and scale data analytics
3. Invest in New Capabilities and the Workforce: investment in technologies such as neuromorphic computing must be made energy efficient and extensive workforce training is required to remain competitive

“It is imperative for the United States to expand and enhance the national research ecosystem and empower researchers with the computational tools so that they can unlock new insights hidden in data,” states Tourassi.