National Strategic Computing Initiative Update
Authors: Dr. William Harrod (US Department of Energy)
BP
Abstract: The National Strategic Computing Initiative (NSCI) focuses the efforts of key US federal agencies into a unified strategy, with five key objectives, to ensure development and deployment of future generations of high-performance computing (HPC) systems. As the reliable progress in computer performance afforded by Moore’s Law approaches its end, this initiative must investigate technologies and approaches for providing new sustainable pathways of both increased capability and capacity. The presenters in this session will describe their agency’s plans and progress towards achieving NSCI’s objectives. Discussions will contribute to furthering NSCI investment plans and identify opportunities for future global collaboration.
Long Description: High-performance computing (HPC) is essential to the global economy, scientific discoveries, and the quality of life worldwide. Over the past sixty years, Moore’s Law in conjunction with Dennard Scaling has provided the computing industry with steady advances in computing performance. Now, as Dennard Scaling has ended, and Moore’s Law is approaching an end within the coming decade, the computing world has become far more complex and must deal not only with an ever-growing demand for increasing compute performance, but also with a new era of explosive data generation and proliferation in all areas of modern society. Furthermore, our ability to derive knowledge from data and computations can be dramatically improved through new sensors and instruments, data science, higher-speed networking, more capable HPC hardware and software, and new mathematical and scientific models, but these will all contribute even more complexity to the current situation.
The National Strategic Computing Initiative (NSCI) focuses the efforts of key federal agencies into a unified national response to these serious challenges through the following five objectives (which are described below):
1. Capable exascale systems
2. Coherent data analytics and computing
3. Technology beyond Moore’s Law
4. HPC ecosystems
5. Public-private partnerships
Accelerate development and deployment of capable exascale systems: The goal is to deploy next-generation exascale systems, including applications and software, by the early 2020s. The key challenges for all exascale systems are: parallelism, system reliability, energy efficiency, and memory and storage. Additionally, significant advances are required in applied mathematics, computational science and computer science research to achieve the promise of these systems.
Enable both data analytics and computing: As the world faces serious, urgent economic, national security, and quality-of-life threats, the need for HPC systems to address both stressing computations and big data analytics is becoming increasingly critical. As data problems become increasingly complex, scalable, energy-efficient artificial intelligence (AI) algorithms will be needed to drive their analytics.
Beyond Moore’s Law: This area addresses research and technology programs to enable new, future, beyond-exascale systems, including new architectures. They may also include heterogeneous subsystems such as neuromorphic or quantum computing components, as well as non-silicon-based devices (e.g., graphene, superconducting devices, etc.). As these HPC systems become more complex and removed from today’s architectures, significant investments must be made in the enabling, supporting HPC ecosystem.
HPC ecosystems: This includes system software, programming environments, networking, cybersecurity and workflow management systems. The goal is to increase both the capacity and capability of HPC ecosystems to enable ease of use and widespread adoption of HPC systems. AI technologies will become an increasingly important component of these ecosystems, both for data analytics and system maintenance.
Public-private partnerships: Public-private partnerships will need to be developed across missions of scientific discovery, economic competitiveness, and national security to ensure the economic viability and commercial availability of promising research results.
The presenters in this session will describe their agency’s plans and progress towards achieving these NSCI objectives. Discussions will contribute to furthering development of investment plans and identify opportunities for future global collaboration.
Conference Presentation: pdf
Birds of a Feather Index