High-Performance Computing for All

The United States faces a global competitive landscape undergoing radical change, transformed by the digital revolution, globalization, the entry of emerging economies into global commerce, and the growth of global businesses. Many emerging economies seek to follow the path of the world’s innovators. They are adopting innovation-based growth strategies, boosting government R&D, developing research parks and regional centers of innovation, and ramping up the production of scientists and engineers.

As scientific and technical capabilities grow around the world, the United States cannot match the traditional advantages of emerging economies. It cannot compete on low wages, commodity products, standard services, and routine or incremental technology development. Knowledge and technology are increasingly commodities, so rewards do not necessarily go to those who have a great deal of these things. Instead, rewards go to those who know what to do with knowledge and technology once they get it, and who have the infrastructure to move quickly.

These game-changing trends have created an “innovation imperative” for the United States. Its success in large measure will be built not on making small improvements in products and services but by transforming industries; reshaping markets and creating new ones; exploiting the leading edge of technology creation; and fusing diverse knowledge, information, and technology to totally transform products and services.

The future holds unprecedented opportunities for innovation. At least three profound technological revolutions are unfolding. The digital revolution has created disruptive effects and altered every industrial sector, and now biotechnology and nanotechnology promise to do the same. Advances in these fields will increase technological possibilities exponentially, unleashing a flood of innovation and creating new platforms for industries, companies, and markets.

In addition, there is a great and growing need for innovation to solve grand global challenges such as food and water shortages, pandemics, security threats, mitigating climate change, and meeting the global need for cheap, clean energy. For example, the energy and environmental challenges have created a perfect storm for energy innovation. We can move to a new era of technological advances, market opportunity, and industrial transformation. Energy production and energy efficiency innovations are needed in transportation, appliances, green buildings, materials, fuels, power generation, and industrial processes. There are tremendous opportunities in renewable energy production, from utility-scale systems and distributed power to biofuels and appropriate energy solutions for the developing world.

Force multiplier for innovation

Modeling and simulation with high-performance computing (HPC) can be a force multiplier for innovation as we seek to answer these challenges and opportunities. A simple example illustrates this power. Twenty years ago, when Ford Motor Company wanted safety data on its vehicles, it spent $60,000 to slam a vehicle into a wall. Today, many of those frontal crash tests are performed virtually on high-performance computers, at a cost of around $10.

Imagine putting the power and productivity of HPC into the hands of all U.S. producers, innovators, and entrepreneurs as they pursue innovations in the game-changing field of nanotechnology. The potential exists to revolutionize the production of virtually every human-made object, from vehicles to electronics to medical technology, with low-volume manufacturing that could custom-fit products for every conceivable use. Imagine the world’s scientists, engineers, and designers seeking solutions to global challenges with modeling, simulation, and visualization tools that can speed the exploration of radical new ways to understand and enhance the natural and built world.

These force-multiplying tools are innovation accelerators that offer an extraordinary opportunity for the United States to design products and services faster, minimize the time to create and test prototypes, streamline production processes, lower the cost of innovation, and develop high-value innovations that would otherwise be impossible.

Supercomputers are transforming the very nature of biomedical research and innovation, from a science that relies primarily on observation to a science that relies on HPC to achieve previously impossible quantitative results. For example, nearly every mental disease, including those such as Alzheimer’s, schizophrenia, and manic-depressive disorders, in one way or another involves chemical imbalances at the synapses that cause disorders in synaptic transmission. Researchers at the Salk Institute are using supercomputers to investigate how synapses work (see http://www.compete.org/publications/detail/503/breakthroughs-in-brain-research-with-high-performance-computing/). These scientists have tools that could run a computer model to produce a million different simulations to produce an extremely accurate picture of how the brain works at the molecular level. Their work may open up pathways for new drug treatments.

Farmers around the world need plant varieties that can withstand drought, floods, diseases, and insects, and many farmers are shifting to crops tailored for biofuels production. To help meet these needs, researchers at DuPont’s Pioneer Hi-Bred are conducting leading-edge research into plant genetics to create improved seeds (see http://www.compete.org/publications/detail/683/pioneer-is-seeding the-future-with-high-performance-computing/). But conducting experiments to determine how new hybrid seeds perform can often take years of study and thousands of experiments conducted under different farm management conditions. Using HPC, Pioneer Hi-Bred researchers can work with astronomical numbers of gene combinations and manage and analyze massive amounts of molecular, plant, environmental, and farm management data. HPC speeds up obtaining answers to research problems from times of days and weeks to a matter of hours. HPC has enabled Pioneer Hi-Bred to operate a breeding program that is 10 to 50 times bigger than what would be possible without HPC, helping the company better meet some of the world’s most pressing needs for food, feed, fuel, and materials.

Medrad, a provider of drug delivery systems, magnetic resonance imaging accessories, and catheters, purchased patents for a promising interventional catheter device to mechanically remove blood clots associated with a stroke (see http://www.compete.org/publications/detail/497/high-performance-computing-helps-create-new-treatment-for-stroke-victims/). But before starting expensive product development activities, they needed to determine whether this new technology was even feasible. In the past, they might have made bench-top models, testing each one in trial conditions, and then moved to animal and human testing. But this approach would not efficiently capture the complicated interaction between blood cells, vessel walls, the clot, and the device. Using HPC, Medrad simulated the process of the catheter destroying the clots, adjusting parameters again and again to ensure that the phenomenon was repeatable, thus validating that the device worked. They were able to look at multiple iterations of different design parameters without building physical prototypes. HPC saved 8 to 10 months in the R&D process.

Designing a new golf club at PING (a manufacturer of high-end golf equipment) was a cumbersome trial-and-error process (see http://www.compete.org/publications/detail/684/ping-scores-a-hole-in-one-with-high-performance-computing/). An idea would be made into a physical prototype, which could take four to five weeks and cost tens of thousands of dollars. Testing might take another two to three weeks and, if a prototype failed to pass muster, testing was repeated with a new design to the tune of another $20,000 to $30,000 and six more weeks. In 2005, PING was using desktop workstations to simulate some prototypes. But one simulation took 10 hours; testing seven variations took 70 hours. PING discovered that a state-of-the-art supercomputer with advanced physics simulation software could run one simulation in 20 minutes. With HPC, PING can simulate what happens to the club and the golf ball when the two collide and what happens if different materials are used in the club. PING can even simulate materials that don’t currently exist. Tests that previously took months are now completed in under a week. Thanks to HPC, PING has accelerated its time to market for new products by an order of magnitude, an important benefit for a company that derives 85% of its income from new offerings. Design cycle times have been cut from 18 to 24 months to 8 to 9 months, and the company can produce five times more products for the market, with the same staff, factory, and equipment.

At Goodyear, optimizing the design of an all-season tire is a complex process. The tire has to perform on dry, wet, icy, or snowy surfaces, and perform well in terms of tread wear, noise, and handling (see http://www.compete.org/publications/detail/685/goodyear-puts-the-rubber-to-the-road-with-high-performance-computing/). Traditionally, the company would build physical prototypes and then subject them to extensive environmental testing. Some tests, such as tread wear, can take four to six months to get representative results. With HPC, Goodyear reduced key product design time from three years to less than one. Spending on tire building and testing dropped from 40% of the company’s research, design, engineering, and quality budget to 15%.

Imagine what we could do if we could achieve these kinds of results throughout our research, service, and industrial enterprise. Unfortunately, we have only scratched the surface in harnessing HPC, modeling, and simulation, which remain largely the tools of big companies and researchers. Although we have world-class government and university-based HPC users, there are relatively few experienced HPC users in U.S. industry, and many businesses don’t use it at all. We need to drive HPC, modeling, and simulation throughout the supply chain and put these powerful tools into the hands of companies of all sizes, entrepreneurs, and inventors, to transform what they do.

Competing with computing

The United States can take steps to advance the development and deployment of HPC, modeling, and simulation. First, there must be sustained federal funding for HPC, modeling, and simulation research and its application across science, technology, and industrial fields. At the same time, the government must coordinate agency efforts and work toward a more technologically balanced program across Department of Energy labs, National Science Foundation–funded supercomputing centers, the Department of Defense, and universities.

Second, the nation needs to develop and use HPC, modeling, and simulation in visionary large-scale multidisciplinary activities. Traditionally, much federal R&D funding goes to individual researchers or small single-discipline groups. However, many of today’s research and innovation challenges are complex and cut across disciplinary fields. For example, the Salk Institute’s research on synapses brought together anatomical, physiological, and biochemical data, and drew conclusions that would not be readily apparent if these and other related disciplines were studied on their own. No matter how excellent they may be, small single-discipline R&D projects lack the scale and scope needed for many of today’s research challenges and opportunities for innovation.

Increasing multidisciplinary research within the academic community should overcome a host of barriers such as single-discipline organizational structures; dysfunctional reward systems; a dearth of academic researchers collaborating with disciplines other than their own; the relatively small size of most grants; and traditional peer review, publication practices, and career paths within academia. Federal policy and funding practices can be used as levers to increase multidisciplinary research in the development and application of HPC, modeling, and simulation.

Third, the difficulty of using HPC, modeling, and simulation tools inhibits the number of users in academia, industry, and government. And since the user base is currently small, there is little incentive for the private sector to create simpler tools that could be used more widely. The HPC, modeling, and simulation community, including federal agencies that support HPC development, should work to create better software tools. Advances in visualization also would help users make better use of scientific and other valuable data. As challenges and the technologies to solve them become more complex, there is greater need for better ways to visualize, understand, manage, monitor, and evaluate this complexity.

Fourth, getting better tools is only half the challenge; these tools have to be put into the hands of U.S. innovators. The federal government should establish and support an HPC center or program dedicated solely to assisting U.S. industry partners in addressing their research and innovation needs that could be met with modeling, simulation, and advanced computation. The United States should establish advanced computing service centers to serve each of the 50 states to assist researchers and innovators with HPC adoption.

In addition, the nation’s chief executives in manufacturing firms of all sizes need information to help them better understand the benefits of HPC. A first step would be to convene a summit of chief executive officers and chief technical officers from the nation’s manufacturing base, along with U.S. experts in HPC hardware and software, to better frame and address the issues surrounding the development and widespread deployment of HPC for industrial innovation and next-generation manufacturing.

If it takes these steps, the United States will be far better positioned to exploit the scientific and technological breakthroughs of the future and to fuel an age of innovation that will bring enormous economic and social benefits.