 |
 |
|
||||||||||||||
 | ||||||||||||||
|
Special Interview: Taking on the Challenge of a 10-Petaflop Computer
|
|||
| Target: World's fastest at 10 petaflops | |||
 | |||
|
----What will Japan's next-generation computer be like? Himeno: It will be a general-purpose supercomputer. When completed, it will become the world's fastest with a computing speed of 10 petaflops (10 to the power of 16 operations per second). The supercomputer will be completed in March 2012. ----As of November 2005, the U.S. boasts the world's fastest computer, known as "BlueGene/L," with a computing speed of 280.6 teraflops (1 tera =10 to the power of 12). Why did you set the target speed at 10 petaflops, which is about 50 times faster than the current fastest supercomputer? Himeno: Based on past transitions in computing speed, a 2- to 3-petaflop supercomputer is certainly achievable. Our 10-petaflop supercomputer will become the fastest when completed in 2012. That is one reason. However, our goal is not to establish a new record, which will only be temporary. Rather, our goal is to develop a new computer that will enable us to do computations that have not been done before, and to contribute to achieving breakthrough results. We decided on a speed of 10 petaflops, taking these goals into account. ----Why are you starting the project at this time? Himeno: Four years have passed since the Japan Agency for Marine-Earth Science and Technology (JAMSTEC) started to operate their "Earth Simulator" (35.8 teraflops) in Yokohama, and at least five years are necessary to develop a supercomputer from the planning stage to final completion. Advances in computer technology are now so phenomenal that if we fail to develop a new computer in the ten years after the current model was developed, we won't be able to keep up with technological advancements. I am afraid we have no time to waste. ----What is the background that has lead to the development of the next-generation supercomputer? Himeno: I had the chance to put forward my idea of developing a 10-petaflop supercomputer at the RIKEN Scientists Committee* early in 2005. "On a 10-petaflop computer, we will be able to simulate the phenomenon of life itself," I said enthusiastically. However, most of the biologists showed a very negative attitude. It seemed to me that they looked on a computer scientist like me as a heretic, because biologists generally base their research activities on experiments. However, one reason I admire RIKEN is because all the scientists are realistic and sensible. In the scientific world, what to a handful of scientists often starts off as a belief, although often doubted by the majority, may lead to new discoveries. For this reason, scientists at RIKEN do not actively try to stop a project going ahead, even when the majority of scientists are opposed to it. Following the advice of Dr. Koji Kaya, Chairman and Director of the RIKEN Discovery Research Institute at Wako Institute, we formed a working group for further discussions, and we used the results for continued discussions at the RIKEN Scientists Committee. In July 2005, we submitted a proposal named "On the Necessity of a Petaflop Computer and an Approach to Research and Development" to the Board of Directors, which accepted the proposal and made the decision to go ahead. RIKEN was selected as the main research body in October 2005 by the Ministry of Education, Culture, Sports, Science and Technology, which was then considering developing a next-generation supercomputer. ----Why was RIKEN selected? Himeno: The computer itself is only a tool. The important thing is how we use it in what fields of science. We have a wide range of needs because RIKEN is constituted by a core of many scientists with a wide variety of scientific backgrounds. As in the past, RIKEN has continued to develop special-purpose computer systems for molecular dynamics simulations to calculate the forces acting between molecules and atoms. "MDGRAPE-3," a special-purpose computer under development by Mokoto Taiji's team at the RIKEN Genomic Sciences Center, will be completed soon, boasting the world's fastest speed of 1 petaflop. RIKEN also has a good track record, having developed the world's first computer system combining scalar (serial processing), vector (parallel processing), and special-purpose machines. I think RIKEN was approved because it has the technical capability needed to develop the next generation of supercomputers and is capable of supporting applications across a broad range of areas. Past successful accomplishments with big projects such as SPring-8 must have contributed significantly to the approval. |
|||
| What can be achieved with a 10-petaflop computer? | |||
| |||
|
----What can be done with the 10-petaflop computer once it is completed? Himeno: In the present circumstances, it is somewhat far-fetched to say that the phenomenon of life itself can be simulated on a computer. However, a 10-petaflop computer will allow us to start simulations. Inter-atomic forces govern all life phenomena including reactions between proteins and molecules, the function of membrane proteins to select and take in specific molecules into cells, DNA replication, transcription to RNA, and repair of damaged DNA. Accordingly, the calculation of inter-atomic forces can lead to simulation of the phenomenon of life itself. A 10-petaflop computer will be able to deal with one million atoms, which, over a calculation time of 24 hours, will allow us to simulate protein reactions for 20 nanoseconds. In one month, we will be able to observe a complete protein reaction. We are developing a general-purpose machine. Great progress is also expected in various fields of application including geoscience, environmental science, energy, nanoscience, and manufacturing (Figure 1). ----Will the supercomputer be made available to the world of industry? Himeno: It certainly will. I spent 20 years in an automobile company researching simulations of fluid flow, and so I know how simulations contribute to cost reductions and efficient design. I am sure that the 10-petaflop computer will benefit the industrial world. However, I do not want it to be used simply to reduce calculation times. This also applies to scientific research - I want it to be used for applications that cannot be calculated on any other computer. One interesting example is to use the computer to test all the synthetic ratios of every possible material to discover epoch-making materials for fuel cells. We have already received various suggestions from the industrial world. Since we also need technical know-how to make full use of the computer, we plan to provide assistance to users and education for engineers. This is not Japan's final development of a supercomputer. As a matter of fact, a speed of 10 petaflops is too slow to simulate the whole body including tissue, blood flow, and movement. We will have to continue our development towards 100-petaflop, then exaflop machines (one exa = 10 to the power of 18). We also have an obligation to train young engineers and students so that they can develop the next generation of supercomputers. ----What is the structure of the hardware? Himeno: Our proposal is for a machine that combines special-purpose, vector, and scalar machines. A combined machine is an advantage, because each machine can compensate for the weak points of the other machines when a broad range of applications is required. However, consideration of applications, in other words, what we should do with it, takes precedence over the hardware. Late in January 2006, we arranged for representatives of the various research fields in Japan, including physics, chemistry, seismology, biology, space science, and meteorology to get together at a meeting of the Application Working Group. We have received information on many projects that will be important in 2012 and after, that cannot be simulated by current computers, but that can be performed on a 10-petaflop machine. From these projects, we will choose several target projects and determine the structure of the hardware so that it will be able to provide maximum performance for these projects.  |
|||
| The supercomputer will be completed in March 2012 | |||
| |||
----What is the overall schedule?Himeno: To begin with, we will look at the target applications, and carry out conceptual design in fiscal 2006. Then, we will proceed to the detailed design and circuit design. Manufacturing will start in April 2010. Then we will start assembly in the summer, and the computer will be partially completed by March 2011. Partial operation will start from April along with system enhancement. The computer will be completed by March 2012, and will be in full operation from April 2012. What we have to do is to ensure that the computer operates stably, and to show the world's first, epoch-making results within half a year. First of all, we are planning to conduct experimental studies on several subjects in nanoscience and life science. You will see the new results on the covers of Nature and Science. ----Where is the construction site? Himeno: We have not decided yet. In constructing the computer, we will use our ingenuity to make the appearance of the computer outstanding. I have an image in my mind of a spherical computer housing (upper part of cover page, Figure 2). A sphere is the ideal shape that will allow the shortest length for the distribution cables between the individual computers. I would very much like to have the exterior walls made of clear glass. The display for the calculation results should be interesting too. I would like to create a computer center where not only researchers but also adults and children can get together. Science thrives only when it has the support of everyone. We should keep this firmly in mind. Anyone will be able to use the computer from any part of the world through a network, but this center will serve as a research center that scientists from all over the world can visit. Good simulation must be accompanied by good experimental results. Science progresses when researchers engaged in simulation and experiments, and when researchers in other areas try to improve their work by learning from others. So it is essential that researchers have the chance to take part in face-to-face discussions with each other. |
|||
| Dreams give us the power to act | |||
| |||
|
----Your field is fluid flow simulation, isn't it, Dr. Himeno? Himeno: Yes, it is. It was in 1985 that I started simulations of the airflow around a car. In those days, I was the first in the world with whatever I simulated, because no one was doing simulations. It was very interesting. Of course, my initial simulations were simple and primitive. However, the simulations I am carrying out now are very complex. This is the result of following my own dream, and that has lead me to what I am doing now. ----What is the most important thing for the development of next-generation computers, and the generations that will come after that? Himeno: We need to dream. A stand-alone personal computer, or a cluster of computers, is good enough for personal use. But we are developing a 10-petaflop computer because we dream of creating new methods that will accelerate research by leaps and bounds, and to broaden research and development into new areas. I have another dream; I hope that I will witness the elucidation of the mystery of life itself on a computer. It is a great feeling to be the first to do something that no one has else has done before. I would like to share this pleasure with as many scientists as possible. 
|
|||
|
※1:Flops Acronym for Floating-point Operations Per Second. Flops is a measure of a computing system's processing speed. Ten petaflops means 10 quadrillion, or "10 to the power of 16" operations per second.
※2:RIKEN Scientists Committee |
|||
| |||
|
|||
[ Top ] |
A 10-petaflop* supercomputer will be built in 2012. The supercomputer will boost a computing speed that is 50 times faster than the world's current fastest computer. RIKEN is responsible for the development, construction, and operation of the supercomputer, which is a huge government project with a total budget of 110 billion yen including facilities for the system and the research grid project. We interviewed Dr. Ryutaro Himeno, Development Group Director at the newly-established Next-Generation Supercomputer R&D Center to ask about why a speed of 10 petaflops, and what they are hoping to achieve with a 10-petaflop supercomputer.|
Contributing to Solving Food Problems with Safe Pesticides
|
|||
|
|||
| Items that have long been used safely are safe in the truest sense | |||
|
"In a study to determine the antifungal effect of a bitter ingredient extracted from citrus fruit, I dissolved it in the presence of sodium bicarbonate. Surprisingly, the sodium bicarbonate, not the bitter ingredient, was found to be highly effective. This led to the start of my current research," says Arimoto, looking back on events that happened some thirty years ago. Sodium bicarbonate is a compound that has long been used as a major ingredient in baking powder and stomachics. "We had been thinking about the safety of pesticides. In developing a new pesticide, very rigorous inspection is performed to assess its toxicity and environmental impacts. In this context, some problems can arise unexpectedly, including the possible action of certain compounds as endocrine disruptors (commonly known as environmental hormones). Such adverse effects will be revealed only after the pesticide has been in use for a long time. The conclusion we had reached on the pesticide safety issues after a long course of investigation was that items that have long been used safely are safe in the truest sense. They are represented by substances that have long been used safely as foodstuff or food additives. We designated the concept of using them for pest control using the acronym SaFE (Safe and Friendly to Environment), and began developing a new pesticide." |
|||
| Discovering a new coating agent | |||
|
Arimoto and others developed a chemical agent based on sodium bicarbonate and acquired official approval for its registration as an agricultural pesticide in 1982. However, that was only the beginning of their hard work. Although the new pesticide was highly effective in controlling powdery mildew, a major disease affecting fruit trees and vegetables, it also produced crop damage. "When the pesticide was applied to crops in a dry greenhouse, the water in the pesticide evaporated, allowing the crystallization of the active ingredient sodium bicarbonate. The resulting crystals dissolved in the night dew to produce a very dense solution, which in turn damaged the crops." Later, potassium bicarbonate (potassium hydrogen carbonate), a similar substitute, but having potassium in place of sodium, was found to be as effective as sodium bicarbonate in controlling the disease. Potassium bicarbonate is also used as a food additive in Europe and the United States. In Japan, it has a long history of use in pharmaceuticals, fertilizers and the like. "In an attempt to suppress the crystallization to prevent chemical damage, I performed many experiments with sodium bicarbonate and potassium bicarbonate in combination with various other chemicals." Arimoto and others tested about 500 to 600 different chemicals, but after nearly four years of research none proved to be satisfactory. "As might been expected, I was about to give up the study. Then, when I was dispensing sodium bicarbonate into test tubes, one tube was left over. Placing my last hope on the experiment, I mixed the tube of sodium bicarbonate with a chemical agent called a glycerol fatty acid ester, and that gave great results!" Dissolving this mixture in water resulted in the formation of droplets about 0.1 to 0.03 mm in diameter. The droplets comprised high concentrations of sodium bicarbonate entrapped in a glycerol fatty acid ester membrane; the same phenomenon was observed with potassium bicarbonate (background image in lower panel on the cover page). "Traditionally, 200-fold dilutions in water have been effective in controlling the causal fungus. But now, even 1000-fold dilutions are effective. As a result of the significant reduction in the concentration in the entire solution, the active ingredient potassium bicarbonate has become detoxifiable by the plant, which in turn has dramatically reduced chemical damage. Because the potassium bicarbonate concentrations in the localized droplets are essentially very high, however, they are considered to be adequately effective against the fungus. In applications to foodstuff and food additives, the agent's control effect is poor when it is used as is. Therefore, this coating technique and other formulation methods are of great importance." Arimoto and others developed a chemical agent based on potassium bicarbonate as the active ingredient. In 1993, the new chemical gained approval for registration as an agricultural pesticide, and was launched under the trade name of 'Kalligreen' (Toagosei Co., Ltd.). This pesticide also serves as a fertilizer, enhances the disease resistance of plants, and has a remedial effect. Currently, Kalligreen is used in more than 90% of vineyards where grapes for Californian 'organic wine' are grown. "California is famous for its strict environmental standards, including regulations on automobile exhausts. With its high safety, Kalligreen has cleared the State's severe requirements for organic cultures." Kalligreen is used widely all over the world; in Japan, it is used mainly for controlling pests on strawberries and cucumbers. Kalligreen represents a good example of the actual application of the SaFE concept worldwide, in which foodstuff and food additives are used to control pests that affect crops. |
|||
| Overcoming the pesticide resistance issues | |||
"I was deeply moved when I held the first commercial pack of Kalligreen in my hand. However, a question soon started to haunt me: what is the benefit of our new pesticide developed based on the SaFE concept? Since there are many pests that cannot be controlled with Kalligreen, conventional pesticides have to be used in combination. Our efforts are meaningless unless all kinds of pests can be controlled with safe pesticides. We began studying how to control other major pests."With its 'body' outside the plant, the fungus that causes powdery mildew thrusts its 'root' into the plant to absorb nutrients for its growth. Kalligreen exhibits its effect by acting on the 'body' of the fungus outside the plant. It should be noted, however, that Kalligreen is ineffective against most fungi that produce crop damage because they grow inside the plant. "Then, I searched for other pests whose 'body' is outside the plant, and selected mites, a major pest on many crops, as the target of the study. Mites insert their 'mouth' in the plant from outside and suck out the nutrients. I realized that this occurred in just the same way as the fungus that causes powdery mildew. Everyone laughed at me, but I was able to have this bold idea since I am not a pathologist or a entomologist." Arimoto and others examined as many as 1000 kinds of foodstuffs, food additives, and related compounds. They applied each test material to mites and determined their effects. Eventually, they found that the propylene glycol fatty acid ester, a compound used as a foam retainer for cakes, was effective. They proceeded to develop a new chemical agent with this compound as the active ingredient. In 2001, the chemical gained approval for registration as an agricultural pesticide, and was launched under the trade name of Acaritouch (Toagosei Co., Ltd.). As such, the new pesticide comprises high concentrations of the propylene glycol fatty acid ester in the form of droplets about 0.05 to 0.03 mm in diameter dispersed in water. These droplets are considered to obstruct the holes through which the mites breathe (spiracles) and suffocate them (Figure 1). "Because conventional chemical pesticides pinpoint and block a particular pathway in the metabolic system, pests that can detour this pathway emerge soon after their use. Miticides encounter the emergence of resistant mites and become ineffective in three to five years after launch. Mites cannot be controlled with any currently-available commercial pesticide, with the only exception of Acaritouch, which is effective because it suffocates them." Kalligreen has not encountered resistance issues. "Following Kalligreen, two kinds of chemical pesticides with excellent efficacy were launched, but both gradually became ineffective within several years due to the emergence of resistant strains. Since the launch of Kalligreen in 1993, no resistance issues have been reported in these 13 years." Regarding the mechanisms of action of Kalligreen, it is postulated that the absence of resistance is thanks to the fact that it breaks the ionic balance in the fungal body to influence the entire metabolic system of the fungus with droplets of high concentrations of potassium bicarbonate, rather than pinpointing inhibitions in the metabolic system. |
|||
| Why do plants get diseased? | |||
|
While continuing to develop new pesticides based on the SaFE concept, Arimoto and others are working to elucidate the mechanisms behind pathogenesis in plants. "There are about 60,000 kinds of mold, scientifically known as filamentous fungi. For example, rice is affected by about 60 fungal species, of which only a few cause significant damage and need to be controlled. Why are only limited fungi capable of entering particular plants to cause disease? I want to know the answer." Arimoto noted the mechanisms that Penicillium digitatum (the microorganism that causes common green mold in citrus fruit) uses to enter mandarin orange cells. Although this fungus is capable of feeding on dead cells of various plants, the disease does not develop in any living cells, except in the mandarin orange after its peel has been damaged. "I had been thinking day and night about how to solve this riddle. However, I was not able to find even a clue. On one occasion, I intentionally damaged the peel and then washed it with water, in an attempt to remove the ingredient secreted from the peel. Surprisingly, the disease did not develop." This demonstrated that the mandarin orange peel did not serve as a barrier to preventing the entry of the fungus, and this discovery in turn led to the elucidation of the following onset process by Arimoto and others. This fungus cannot digest living cells for its own nutrition. When damaged, the peel secretes an oily substance, which kills the cells inside the fruit and renders them digestible for nutrition by the fungus. "However, provided that the oily substance is washed off immediately, the cells inside the fruit remain alive, in which condition the fungus cannot digest them, and so that the disease does not develop." While the fungus is digesting mandarin orange cells for its own nutrition, an acidic substance is produced, resulting in a decrease in the hydrogen ion concentration to a pH of 3.5. At this level, the nature of the fungus changes and it becomes able to bore into the walls of the surrounding living cells and enter the cells. This is how the fungus enters the fruit and causes common green mold. "When the pH level has dropped to 3.5, the fungus's nature changes dramatically; the fungus becomes aggressive towards living cells and begins to enter them. This is just like the Wolfman, who metamorphoses into a wolf when seeing the full moon. In plants other than the mandarin orange, however, no acidic substance is produced, and so that the pH level does not drop to 3.5. For this reason, common green citrus mold never develops in living plants other than the mandarin orange. What, then, is the difference between living and dead cells? This remains a riddle." Furthermore, Arimoto identified the reason why Fusarium oxysporum (the fungus that causes Fusarium Wilt in tomatoes) does not enter through the leaves, despite it being able to enter the damaged roots and stems of tomatoes. "The glutamine concentration proved to be the key to the metamorphoses of the fungus into the Wolfman. The glutamine concentration is quite high in tomato roots and stems, but is low in the leaves, and so the fungus cannot enter (Figure 2). In plants other than tomatoes, Fusarium Wilt does not develop because of the very low glutamine concentrations. It is postulated that when the glutamine concentration exceeds a particular level, an unidentified gene is turned on in the fungus to start entry." Finally, Arimoto talked about his future goals. "The mechanisms of the onset of disease in plants are quite difficult to elucidate, and this represents a research theme for which almost no achievements have been registered to date. This research theme can only be investigated by an organization like RIKEN's Discovery Research Institute, where researchers can dedicate themselves to pursuing solutions to genuine questions without hurrying. Elucidating the onset mechanisms would lead to the development of safer, more effective, groundbreaking methods for controlling disease. In addition to developing more safe pesticides based on the SaFE concept, I want to uncover a new hint to help elucidate the mechanisms of the onset of plant disease, and open the frontiers that will attract many researchers." 
 |
|||
|
|||
 | |||
|
|||
[ Top ] |
The world's population now exceeds 6.5 billion people and is even increasing dramatically, whereas the expansion of croplands is subject to limitations. In addition, there is general concern about reductions in crop yields due to soil loss, depletion of underground water for irrigation, desertification, urbanization, climate change and the like. To feed the people of the world with the limited availability of croplands, it is necessary to efficiently protect for crops against weeds, insects and disease. Currently, about 30% of crop yields are lost due to damage by such pests. Without the use of pesticides, the current yield levels would decrease by as much as one half. It should be noted, however, that pesticides must be safe. "No compound can be determined to be completely safe. We are carrying out research to find new applications for selected compounds for use as safe pesticides from among existing materials that have long been used safely as foodstuffs or food additives," says Yutaka Arimoto, Leader of the Applied Biology for Plant Protection Research Unit at RIKEN's Discovery Research Institute. He and others are working to develop safe pesticides and to elucidate the mechanisms behind pathogenesis in plants, in an attempt to contribute to ensuring food safety.