George Michaels
High-Performance Computing
June / July 2005
Volume 3 Number 3
We can only see a short ways ahead, but we can see plenty there that needs to be done.
—€”Alan Turing (1912—€“1954)
Scientific research today involves investigating how whole systems work together rather than studying their disparate parts. Scientists seek to understand how an organism or system functions so that they can predict and prevent disease or disaster, rather than try to remediate the problem afterward. This approach demands a research environment equipped with powerful computational tools for simulating and modeling across multiple disciplines.
This research approach also generates massive amounts of data that must be stored, organized, analyzed and processed into usable information. It is a task that defeats current computing technologies. Moreover, the computational environments necessary to solve the multi-scale challenges of systems science research are at least three to five years from becoming available.
Monumental challenge
Many innovative computational advances in use now were developed in support of the Human Genome Project, the first major data-intensive computing challenge.
But these tera-scale computational platforms are not adequate to support the challenges of post-genome data analysis of proteomics research. Advancements, such as the ability to study the amount of each protein present at any time, are essential as scientists attempt to learn the role of proteins in important cellular functions and communications.
Experts predict that the potential data production of proteomic research will be roughly four orders of magnitude greater than for the Human Genome Project.
That's an even bigger jump than the gap between the Wright brothers' plane the space shuttle.
The high performance computing and scientific communities face an enormous challenge that undoubtedly will require a paradigm shift from simply building bigger, faster computers to an entirely new way at looking at the problem. It is truly a grand challenge to create peta-scale computational, data handling, and modeling and simulation architectures and capabilities. In many cases, the challenge also includes developing scalable data management and analysis methods that address—€”in real time—€”he complexities of managing and fusing data in disparate formats and multiple modalities such as sensor, text, audio and video data.
National need
Dealing with the exponential growth in the amount of data collected from experiments, measure and observation is one of today's most urgent technical challenges. Government research laboratories and industry leaders are working together to drive the development of new computing paradigms in both supercomputing architectures and scalable software.
The need is real. For example, real-time analysis is critical to managing the next generation of this nation's power grid, meeting the intelligence analysis demands to ensure national security, driving scientific development of new energy sources, solving the problems of climate and environment and delivering personalized medical care.
Federal support
Federal research initiatives have played a critical role in keeping the United States at the cutting edge of science and computing. Initiatives include the Next-Generation Internet, the Department of Energy's DOE 2000 Project, and the National Science Foundation's Knowledge and Distributed Intelligence emphasis.
It has been federal support that has allowed the research community to "live in the future" and to tackle long-term, high-risk research challenges.
Federal support for America's national laboratories enables them to play a central research role in advancing high performance computing and developing new algorithmic solutions and scalable software for multi-scale simulations, as well as analytics and data management. The national laboratories are conducting research and development to meet the demands of data-intensive computing challenges while also sending science and technology grounded in advanced computing out to the industrial arena.
For example, new visual analytics tools that display massive quantities of data in visual formats and graphical representations are being used to identify themes, trends, and relationships that otherwise might remain buried in the mountains of text, video or audio files. These tools already have proved valuable for gleaning knowledge from huge data collections within biology and medicine, homeland security, law enforcement or business competitiveness.
Regardless of how advanced a computing system is, protecting government, personal and business information and ensuring that it can't be accessed or compromised by competitors, hackers or terrorists can't be left to add-on technologies. Next-generation high performance computing infrastructures must have integrity and security built directly into the computer systems, networks and applications, creating a critical infrastructure that is intrinsically secure.
In the process of scientific discovery, computational power, combined with next-generation abilities to model and simulate, has become an equal partner with theory and experiment. Further investments in next-generation, high performance computing platforms, large-scale optical networks, plus collaboration, simulation and modeling software are essential for enabling researchers to achieve a whole new understanding of our world through systems science.
George Michaels is Associate Laboratory Director for Computational and Information Sciences at Pacific Northwest National Laboratory.
Copyright © 2012 | Innovation America