This has been taken from the C3.ca Innovation Strategy proposal.
What is High-performance Computing?
Traditionally, High Performance Computing (HPC) has been synonymous with supercomputing. Loosely speaking, supercomputing can be defined as a computing infrastructure that is at least 100 times more powerful than a state-of-the-art desktop machine and might be, by today's standard, tens of thousands times more powerful. Powerful is the operative word. The most powerful desktop PC will always be a lightweight when compared to state of the art HPC. A computation that might take a year on a desktop machine reduces to "a coffee break" on a world leading HPC machine. Research productivity, time to manufacture and market, information acquisition, knowledge discovery and innovation are significantly enhanced.
With the rapid evolution in HPC, the universe of applications has exploded and the technological definition of supercomputing has proved too narrow. High-performance computing is no longer just a piece of equipment you buy. It is a rapidly evolving discipline focused on the ability of new generations of information technology, sharing large-scale computing power, processing huge amounts of shared data, in a national grid environment. The resulting leap in computing capability supports new methods of computer simulation, modelling and visualization in artificial reality environments. Consequently, many more elements than a supercomputer are needed to define HPC.
HPC embraces large capability computers, massive memory to handle specialized data and unique software developed by highly skilled people. It utilizes the highest capacity communications networks to exchange and retrieve enormous data sets with results frequently delivered to HPC driven visualization stations or other specialized output devices.
High Performance Computing is a core research infrastructure in the
knowledge economy.
Where several years ago these supercomputers were considered only for
limited usage, it is now considered an "entry ticket" for organizations
conducting research in most fields where Canada wants to innovate.
Why Invest in High-performance Computing?
The advent of the computer has revolutionized research methods and is helping to map new paths to technological innovation. It has created a third type of scientist - the computational scientist - who joins the traditional theorists and experimentalists. A computational scientist uses the computer as a tool to build mathematical models and simulate the behaviour of complex systems, which range from galaxies to thunderstorms to molecules to airplanes. Using HPC, the experimental scientist simulates the systems being studied in the laboratory to help identify the right new experiments to do, opening up entirely new modes of thought, and providing deeper insight into the systems under study.
Consider these examples:
Human Genome
Genome Canada is making great strides in the emerging field of genomics. Leaders in this field are critically dependent on HPC. Why?
"The human body is undoubtedly the most complex machine ever created.
Genome researchers are undertaking the challenging task of unraveling
how it is organized and how it functions.
High-performance computing plays a dominant role in this research.
Without extremely sophisticated computational infrastructure,
Genomics research would not be possible."
(Professor Christoph Sensen, University of Calgary)
The human genome has been mapped with the support of enormous computing resources. This is an incredible scientific accomplishment for we now possess the genetic equivalent of an alphabet and sentences. But these text segments must now be assembled into the book of life. As with learning any new language, interpretation is the more considerable challenge and requires another huge increase in resources. There are more supercomputers in the world working on this scientific problem than on any other. The scientific findings, of course, create the basis for the prevention and/or cure of disease and genetic problems.
Other nations are accelerating their research. By continuing to invest in exploring the human genome, Canada will emerge as the world leader in a number of critical areas in genomics and can retain its high standing in the world, second only to the United States.
Proteomics
Who says life is simple? The availability of the human and other genome sequences, with many billions of base-pairs of DNA, has opened up the possibility of studying the protein composition of normal and aberrant cells at the whole cell level. This has led to the creation of the science of proteomics. Proteins, which carry out most of the biochemical processes in our bodies, are used as structural material (muscles), enzymes (chemistry), receptors, antibodies, and peptides. Proteomics is the study of the entire collection of proteins of a cell, tissue or organism. The computational tasks are so large; HPC is a hugely critical enabling technology. Why?
A single human cell may contain over 1 million different proteins if biologically important states of post-translation modification are taken into account. The computational biology (high performance computing) associated with proteomics can be divided into two broad areas. The first is bioinformatics, which can be defined as data analysis and data acquisition. Bioinformatics includes the development of more sophisticated data submission and data visualization tools, the maintenance and annotation of databases, the development of data mining software for pattern finding and the creation of text mining software for data extraction. The second broad area is simulation which can be defined as the creation of visually and informationally rich "movies" of cellular activity as well as numerical or symbolic analysis for cellular network simulation, whole-cell metabolic simulation, 3D visualization and continuum modeling; macromolecular modelling and dynamics, supramolecular modeling, metabolic modeling and cellular response modeling.
"High performance computing is critical to the advancement of nearly all
aspects of proteomics. Proteins are infinitely more complex than genes
and consequently the computer requirements for proteomics far outstrip
the computer needs for genomics research. High performance computing
is needed for high throughput mass spectral analysis (MDS Proteomics has
"maxed out" their 200 node cluster for more than a year), for structural
proteomics (i.e. protein fold prediction, high throughput homology
modeling), for functional proteomics (analyzing or predicting protein-
protein interactions) and for proteome annotation (which requires large-
scale data searches and data mining). Many of these computations easily
bring large clusters of computers to a near stand-still."
(Associate Professor David Wishart,
Bristol-Myers Squibb Chair, University of Alberta)
Weather and Environmental Prediction
Canada is a world leader in weather modeling. To build on this position, Canadian weather research teams need more and better HPC in order to analyze large amounts of data. Why?
"Improving the accuracy of weather models requires better space and time
resolution and thus more HPC.
More accurate weather modeling, in turn, generates significant economic
benefits - in agriculture, tourism, environmental management and long-term
climate modeling."
(Dr. Jim Abraham, Research Director, Meteorological Services Canada)
Weather and Environmental Prediction in Canada is part of a cooperative international program that monitors and predicts changes in the global atmosphere (weather, climate, and atmospheric chemistry), hydrosphere (rivers, lakes, oceans) and the cryosphere (snow and ice). The program contributes to scientific understanding of weather and climate systems, atmospheric chemistry and changes in weather, climate, air quality and hydrology. It also assesses their impacts on human and natural ecosystems. In addition it conducts research on strategies for adaptation to these atmospheric changes and their impacts. This program also conducts assessments of the state of atmospheric science in support of policy development.
Using its powerful HPC system, Meteorological Service of Canada provides:
While the above material references federal government activity, there are several leading-edge university research groups and private sector organizations doing climate, weather and atmospheric prediction. They too require advanced computational support.
Nanotechnology
Manipulating matter at the nanometer level (one billionth of a meter, so small that millions of tiny nano-structures would fit on the head of a pin) promises to create a new industrial revolution. Canada is building a position as a major player with initiatives such as the recently created National Institute for Nanotechnology, located at the University of Alberta, which combines the efforts of the National Research Council, the University and academics across the country. HPC is vital to the success of this technology cluster. Why?
"In the long-term nanotechnology is all about computing. It's the only
tool we have to bridge the chasm between theory and experiment for hard
problems that resist analytical treatment. Which of course means almost
all problems, especially in nanotechnology where you start wanting to take
into account all of the individual atoms within a structure."
(Professor Mark Freeman, iCORE Chair, Canada Research Chair,
1999 NSERC E.W.R. Steacie Fellow)
Nanotechnology is a tricky business. Physical experiments at this level of precision are still difficult to do and are expensive. This research can be greatly accelerated using HPC. Scientists model the physics of particle interactions at the nano-leve1 using complex computer codes. This eliminates physical experiments that would waste resources, enabling scientists to pinpoint the ones that are likely to succeed. The potential benefits are enormous - and quite varied. Consider microscopic nano-robots that can be put in the bloodstream to attack cancer cells or a film that is only a few atoms in width that can be used to strengthen a surface (e.g., joint replacement surfaces).
-----
The examples above are but a few that are reliant on HPC. Common to all applications and enabling these research advances are two key ingredients:
Large-scale computing is changing the way research and development is being done in many sectors that are of critical importance to the Canadian economy. Investment in HPC is vital to being part of this change.