Parched West is using up underground water: Study points to grave implications for Western U.S. water supply

Science Daily - Thu, 24/07/2014 - 9:21pm
A new study finds more than 75 percent of the water loss in the drought-stricken Colorado River Basin since late 2004 came from underground resources. The extent of groundwater loss may pose a greater threat to the water supply of the western United States than previously thought.
Categories: Science

Brain's dynamic duel underlies win-win choices

Science Daily - Thu, 24/07/2014 - 9:20pm
People choosing between two or more equally positive outcomes experience paradoxical feelings of pleasure and anxiety, feelings associated with activity in different regions of the brain, according to research. In the study, participants made choices between paired products with different or similar values. Choosing between two items of high value evoked the most positive feelings and the greatest anxiety.
Categories: Science

Microbes make the sake brewery

Science Daily - Thu, 24/07/2014 - 9:20pm
A sake brewery has its own microbial terroir, meaning the microbial populations found on surfaces in the facility resemble those found in the product, creating the final flavor according to research. This is the first time investigators have taken a microbial census of a sake brewery.
Categories: Science

Moose drool inhibits growth of toxic fungus

Science Daily - Thu, 24/07/2014 - 9:20pm
Research shows a surprisingly effective way to fight against a certain species of toxic grass fungus: moose saliva. Inspired by an earlier study that showed that moose grazing and saliva distribution can have a positive effect on plant growth, the research team set out to test an interesting hypothesis -- whether moose saliva may, in fact, "detoxify" the grass before it is eaten.
Categories: Science

Antioxidant biomaterial promotes healing

Science Daily - Thu, 24/07/2014 - 9:19pm
The first-ever inherently antioxidant biomaterial has been created by researcher. It has the potential to prevent failure in medical devices and surgical implants. The lead researcher said the new biomaterial could be used to create scaffolds for tissue engineering, coat or build safer medical devices, promote healing in regenerative medicine, and protect cells, genes, and viruses during drug delivery. He added that the new biomaterial is easy to make and inexpensive.
Categories: Science

Biologist warn of early stages of Earth's sixth mass extinction event

Science Daily - Thu, 24/07/2014 - 9:19pm
The planet's current biodiversity, the product of 3.5 billion years of evolutionary trial and error, is the highest in the history of life. But it may be reaching a tipping point. Scientists caution that the loss and decline of animals is contributing to what appears to be the early days of the planet's sixth mass biological extinction event. Since 1500, more than 320 terrestrial vertebrates have become extinct. Populations of the remaining species show a 25 percent average decline in abundance. The situation is similarly dire for invertebrate animal life.
Categories: Science

New approach to form non-equilibrium structures

Science Daily - Thu, 24/07/2014 - 9:19pm
Researchers get closer to understanding the fundamentals of non-equilibrium, self-assembled structures, unlocking potential in a variety of fields. By injecting energy through oscillations, researchers can force particles to self assemble under non-equilibrium conditions, they report.
Categories: Science

New characteristics of complex oxide surfaces revealed

Science Daily - Thu, 24/07/2014 - 9:19pm
A combination of microscopy and data processing has given researchers an unprecedented look at the surface of a material known for its unusual physical and electrochemical properties.
Categories: Science

Role of media in sharing life events

Science Daily - Thu, 24/07/2014 - 9:18pm
To share is human. And the means to share personal news — good and bad — have exploded over the last decade, particularly social media and texting. But until now, all research about what is known as “social sharing,” or the act of telling others about the important events in our lives, has been restricted to face-to-face interactions.
Categories: Science

Genetic switch discovered that can prevent peripheral vascular disease in mice

Science Daily - Thu, 24/07/2014 - 9:18pm
Millions of people in the United States have a circulatory problem of the legs called peripheral vascular disease. It can be painful and may even require surgery in serious cases. This disease can lead to severe skeletal muscle wasting and, in turn, limb amputation. Scientists have tested a non-surgical preventative treatment in a mouse model of the disease and it was associated with increased blood circulation.
Categories: Science

Humans share fairness concerns with other species

Science Daily - Thu, 24/07/2014 - 9:17pm
Humans aren’t the only species to react strongly to actions they consider unfair. A similar drive for fairness in monkeys and some dogs may offer insight into people’s desire for equity, according to experts.
Categories: Science

Gun violence prevention requires public health approach, experts say

Science Daily - Thu, 24/07/2014 - 9:17pm
Preventing gun violence will require a scientific public health approach and recognition of the limits of predicting individual cases of violence, according to experts.
Categories: Science

See the Moon, Mercury and Venus Before Sunrise Friday - Thu, 24/07/2014 - 9:12pm
Mercury and Venus will star in an early morning meet-up with the moon Friday (July 24). Both planets can be seen low in the east-northeast roughly 45 minutes before sunrise.
Categories: Science

Designing the exascale computers of the future

Kurzweil AI - Thu, 24/07/2014 - 9:10pm

Big computers have come a long way since the Mark I (left) at Harvard in 1944, capable of approximately three calculations per second. Today, the fastest supercomputer, China’s Tianhe-2 (right), runs at 33.86 petaflops (quadrillion floating-point operations per second). (Credit: Harvard University Archives and Jack Dongarra.)

Several groups in Harvard’s School of Engineering and Applied Sciences (SEAS), are contributing to the basic science and engineering that will be needed to create the hardware and software for the next generation of supercomputers.

The goal: an exascale machine, performing at least 1018 (sextillian) operations per second — 30 times faster than the current fastest machine in the world, China’s Tianhe-2, capable of an Rmax* of 33.86 petaflops, or 33.86 quadrillion (1015) floating-point operations per second.

Modeling at the molecular scale

Meanwhile, others at SEAS and elsewhere at Harvard are poised to take advantage of the new computing machinery when it becomes available, applying it in fields such as climatology, materials science, molecular biology, astrophysics, and medicine.

Imagine a heart surgeon operating to repair a blocked coronary artery, says Efthimios Kaxiras, the John Hasbrouck Van Vleck Professor of Pure and Applied Physics. “Someday soon, the surgeon might run a detailed computer simulation of blood flowing through the patient’s arteries, showing how millions of red blood cells jostle and tumble through the small vessels. The simulation would identify the best repair strategy. With a fast enough computer, it could all be done in a few minutes, while the operation is under way.”

As leader of an interdisciplinary research group, he has built just such a simulation of “hemodynamics,” tracing the movements of several million blood cells through the filigree network of coronary arteries that supply oxygen to the heart muscle.

The problem: simulating just one second of that blood flow — the duration of a single human heartbeat — took five hours on one of the world’s fastest supercomputers a few years ago. An exascale computer could handle it, and also enhance the detail and verisimilitude of the models.

In fact, if the simulation could reach down to the molecular level, Kaxiras says, it might directly model the way lipid molecules are deposited on the arterial wall, predicting blockages before they form.

The limits of chip speed

The engine driving these amazing gains in computer performance has been the ability of manufacturers to continually shrink the dimensions of transistors and other microelectronic devices, thereby cramming more of them onto a single chip. But the speed limit for modern computers is now set by power consumption. If all other factors are held constant, the electricity needed to run a processor chip goes up as the cube of the clock rate: doubling the speed brings an eightfold increase in power demand.

SEAS Dean Cherry A. Murray, the John A. and Elizabeth S. Armstrong Professor of Engineering and Applied Sciences and Professor of Physics, points out that high-performance chips are already at or above the 100-watt level. “Go much beyond that,” she says, “and they would melt.”

Since 2005, the main strategy for boosting performance has been to gang together multiple processor “cores” on each chip. The clock rate remains roughly constant, but the total number of operations per second increases if the separate cores can be put to work simultaneously on different parts of the same task. Large systems are assembled from vast numbers of these multicore processors.

When the Kaxiras group’s blood flow study ran on the Blue Gene/P at Jülich, the machine had almost 300,000 cores. The Chinese Tianhe-2 supercomputer has more than 3 million cores. An exascale machine may have hundreds of millions of cores, or possibly as many as a billion, the Harvard scientists say.

Reducing power consumption

In principle, an exascale computer could be built today, using chips like those installed in the Tianhe-2 — just many more of them. But such a monstrous machine would come with a monstrous electric bill: The power demand would be about 500 megawatts, half the output of a nuclear power plant, the scientists calculate. To bring exascale computing into the realm of practicality, energy consumption will have to be reduced by an order of magnitude.

Maximizing computational power while minimizing electrical consumption is one of the principal research aims of David M. Brooks, Haley Family Professor of Computer Science at SEAS. A first step in this direction is knowing where the energy goes. Which parts of a processor chip consume the most energy during various phases of operation?

To help answer this question Brooks and his colleagues have created a series of measurement and simulation tools that document the performance of existing chips and predict the behavior of hypothetical alternatives. In this way they can explore the space of possible processor architectures, asking how design choices affect the energy budget. For example, is it more effective to have many small and simple cores on each chip or a few larger ones? Or some of each?

Brooks’ findings emphasize the need to optimize many design parameters in concert, rather than tuning each factor individually. A common strategy for reducing the power consumption of digital circuits is to lower the operating voltage; but lower voltages also make the circuitry more sensitive to variations between one transistor and another, which can lead to errors.

Furthermore, a surge of power demand in one part of a processor can cause a voltage dip elsewhere, much as the room lights dim momentarily when the refrigerator comes on. Again, the result of such a “voltage emergency” could be an incorrect computation. The traditional way of avoiding such errors is to maintain a margin of safety, so that the weakest transistor on the chip gets adequate voltage even under the worst-case operating conditions.

That strategy may no longer be viable as demands for power efficiency become more stringent. Brooks and his group have been examining an alternative scheme in which errors caused by voltage dips are caught and corrected, while a layer of management software works to minimize the number of voltage emergencies.

Power demands may also be reduced by improvements to the design of individual transistors and the materials from which they are fabricated. Sadasivan Shankar, Program Leader for Materials Design at the Intel Corporation, argues that it’s no longer accurate to speak of silicon integrated circuits. Although silicon remains the substrate on which the circuits are inscribed, dozens of other elements from all regions of the periodic table are now essential constituents.

New materials are needed to maintain the speed and reduce the power consumption of computers (and paradoxically, major computing resources are needed to explore the vast space of possible material compositions).

Processor chips are not the only components for which the tradeoff between speed and power consumption is problematic. The energy efficiency of memory chips has not kept up with that of processors; as a result, exascale machines will probably have to get along with less memory per processor core, or new memory devices will need to be invented. KurzweilAI has covered a variety of research projects to do just that.

The energy cost of moving data from place to place within a computer is also claiming a larger share of the overall energy budget. The constraint on interprocessor communication comes straight from the laws of physics — moving data more than a few feet will inevitably introduce delays.

Supercomputing took a turn in a new direction in 2008 with a machine called Roadrunner, built by IBM and installed at Los Alamos National Laboratory. In addition to 6,480 conventional processor chips, it had 12,960 specialized “coprocessor” chips, which had an unusual provenance. They were originally designed for the Sony Playstation game system, where they were well adapted to certain fairly simple but highly repetitive tasks such as displaying graphics or decompressing video. The coprocessor chips performed those tasks with remarkable speed and energy efficiency, and the aim of Roadrunner was to apply them to algorithms in scientific computing that have a similar repetitive structure.

Cherry Murray observes that there was initially some skepticism about the Roadrunner design: “People worried that all their software would have to be rewritten to work with the new hybrid architecture, including operating systems, libraries of utility programs, and software tools such as language compilers.” Indeed it was a daunting task, but the conversion succeeded, and Roadrunner became the first petascale computer — the first to achieve 1015 operations per second.

Several more large-scale hybrid systems have been built since Roadrunner, mostly with graphics processing units (GPUs) as the coprocessors. GPUs, which evolved from the display-controller chips of desktop PCs, have hundreds or thousands of comparatively simple processing cores, all running in parallel. Hybrid computers that include coprocessors of one kind or another now account for about a third of all the computing capacity on the Top 500 list of supercomputers.

Murray sees this development in the context of a decades-long process of evolution. In the 1980s, supercomputers relied on highly specialized, custom-built hardware, based on technologies quite different from those of mass-produced computers. In the 1990s, the high-performance community shifted to “vanilla” off-the-shelf components when they were found to have a better price-to-performance ratio. Now the trend is back to more specialized technologies again.

Murray expects that heterogeneous architectures will rule scientific computing in the coming years. She cites an example: If you try to model the interactions of an electromagnetic field with solid-state matter, the two parts of the problem differ in character. The electromagnetic field extends continuously throughout space, whereas the solid has the discrete symmetries of a crystal. It would be surprising if one computer architecture were optimal for representing both systems; future computers may well have specialized subsystems tuned for different classes of algorithms.

Tracking motions of biological macromolecules

David E. Shaw and his collaborators used a supercomputer called Anton to model at high resolution the molecular dynamics of protein folding, a very rapid process in which polypeptide chains spontaneously self-assemble into characteristic 3D structures. Here, each experimentally derived protein structure is shown in red, and the results of each simulation in blue. (Credit: Kresten Lindorff-Larsen, Stefano Piana, Ron O. Dror, & David E. Shaw [2011] and Science/AAAS.)

The ultimate in specialization is a computer system designed to solve just one kind of problem. A prominent example is Anton, a fleet of computers designed and built by D. E. Shaw Research for studies of the flexing, twisting, and wiggling motions of proteins and other biological macromolecules. The machine is a computing engine for the molecular dynamics method, which traces the movements of all the atoms in a molecule using a two-step procedure.

Given an initial configuration of the molecule, the first step is to calculate the forces between all pairs of atoms; the second step moves the atoms “a little bit” in response to the forces. Then it’s back to the first step to recalculate the forces based on the new positions, followed by a further small movement of each atom, and so on. To capture the fastest molecular vibrations, the entire computation has to be repeated at intervals of 1 femtosecond, or 10–15 second.

Before Anton, no one had simulated more than 10 microseconds of motion for large molecules of biological significance. With Anton it becomes practical to carry out 1012 iterations of the molecular dynamics algorithm, which amounts to a millisecond of biological time — 100 times the previous record, and long enough to observe key events such as protein folding and enzymatic action.

Internally, Anton relies on a high level of parallelism and also on custom integrated circuits designed for this specific application. Shaw estimates it is “100 times faster than the fastest general-purpose supercomputers — but only for molecular dynamics simulations.”

When you upgrade to a new laptop computer, much of your old software will probably work on the new model, and perhaps run faster. In the world of supercomputing, transitions from one generation to the next are not so effortless. The migration is not just a matter of reinstalling software, or even recompiling it from the source code. Getting the full benefit of a new architecture almost always requires adjustments at the algorithmic level.

Kaxiras’ blood-flow simulation provides an example. The usual approach to fluid dynamics calculations is numerical solution of differential equations, but the Kaxiras group adopted a different scheme, called the lattice Boltzmann method, in which continuous flow is approximated by the movement of “fictitious particles” through a gridlike three-dimensional lattice.

This method was chosen in large part because the computation is readily partitioned among thousands of processors. Events at any one node of the lattice depend only on nearby nodes, reducing the need for data movement. Nevertheless, a recent reexamination of the algorithm showed that communications bandwidth is what limits performance; the full computational capacity cannot be put to use because processors stand idle while waiting for data to be fetched from memory or transmitted from other processors. In adapting the Blue Gene/P code to run on the larger Blue Gene/Q in 2012, overall speed was improved by further limiting the data traffic between nearby nodes of the lattice, even though that means some values have to be recomputed.

Amanda Peters Randles, Ph.D.,  a Lawrence Fellow at Lawrence Livermore National Laboratory, reports that getting the simulation to run efficiently on 300,000 cores took a year and a half, and since then the process of updating and revising has never stopped. Recent changes introduced a different set of software libraries, and she rewrote the entire program in a different programming language (C rather than Fortran).

Parallel computing

Hanspeter Pfister, An Wang Professor of Computer Science and director of IACS, believes the move to exascale will call for fundamentally new programming models. “We’re about to hit the wall,” he says. Even at petascale, only the benchmark programs used to rate and rank the machines run at full speed; other software may reach only 10 percent. Other elements of the software infrastructure — operating systems, file systems, and the “middleware” for connecting to databases and networks — are at their limit.

Many programs now running on highly parallel computers are built on the Message-Passing Interface, or MPI. As the name suggests, this collection of utility programs and protocols allows parallel programs to communicate and synchronize their activities by passing messages. At a lower level, computations within GPU chips are most often managed with a system called CUDA (Compute Unified Device Architecture), which allocates tasks to the thousands of cores and provides analogous facilities for communication and synchronization.

These tools will not scale to the next level, Pfister says. “We can’t be thinking about a billion cores in CUDA. And when the next protocol emerges, I know in my heart it’s not going to be MPI. We’re beyond the human capacity for allocating and optimizing resources.”

The remedy, Pfister suggests, may be “brute force.” Some of the labor of managing concurrent computation should be moved into hardware, which already coordinates multiple “threads” of execution in multicore processors. And some new level of abstraction is needed to relieve the programmer of responsibility for micromanaging parallel processes.

One possible model is the MapReduce protocol adopted by Google for computations on massive data sets. Under MapReduce, tasks are automatically parceled out to processors, then results are gathered up, all without the need to specify where and when each piece of work is done. Existing implementations of MapReduce are not ideal for scientific computing tasks, but the basic principle might be adapted to the needs of science just as GPUs have been. “CUDA meets MapReduce,” Pfister suggests.

The HELIX project tackles the problem of parallel programming from another direction. HELIX is a programming-language compiler that automatically arranges for independent sections of a program to run simultaneously. HELIX is the creation of Simone Campanoni, a postdoctoral fellow at SEAS, who works with Brooks and with Gu-Yeon Wei, Gordon McKay Professor of Electrical Engineering and Computer Science.

Still another tool for parallel computing comes from Leslie G. Valiant, the T. Jefferson Coolidge Professor of Computer Science and Applied Mathematics. Valiant, a Turing Award winner, has created a “bridging model” that might help close the gap between parallel hardware and software.

In sequential computing (that is, with one processor), programmers can generally ignore the idiosyncrasies of a particular machine and think in terms of an idealized or abstract computer, called the von Neumann machine. Valiant proposes an analogous model for parallel computers that allows programmers to ignore details such as the pattern of connections among the processors, and to write programs that are portable among different machines. Companies such as Google and Facebook have recently adopted Valiant’s bulk synchronous model for their most onerous computing tasks.

Hardware and software each have an important role to play in the development of the exascale era. Many believe that “co-design” — bringing hardware engineers together with scientists and programmers — will result in the best match between what users want and what manufacturers can supply. Shaw’s Anton was created by such a collaboration, and the U.S. Department of Energy has established three co-design centers to foster this kind of activity. At SEAS, this type of boundary-blurring collaboration happens every day in the hallways of Maxwell Dworkin and, yes, the basement of the Lyman Lab.

Exascale by 2018?

In 2010, the Advanced Scientific Computing Advisory Committee of the U.S. Department of Energy announced a goal — that a computing platform would be operating at exascale by 2018. The schedule has since slipped a little. Committee member Horst Simon, of Lawrence Berkeley National Laboratory, recently suggested that even 2020 may be too optimistic a target.

Murray argues that exascale machines will surely be built, despite all difficulties, “simply because the need for them is so urgent.” One of the most pressing applications, she says, is “stockpile stewardship,” the program of experiments and computer simulations meant to ensure the continued viability of nuclear weapons, which have now been maintained for more than 20 years without real-world testing.

This program alone, Murray says, is demanding enough to ensure that at least a few exascale machines are built in the United States. She adds that certain other applications, such as climate modeling, might also be regarded as matters that impinge on national or global security.

And what comes after exascale? We can look forward to zettascale (1021) and yottascale (1024).

Then we run out of prefixes.

* Not to be confused with the higher Rpeak value, the theoretical peak performance of the system, sometimes misleadingly quoted as a standard performance measure. See the Top 500 lists for more.

Categories: Science

Science Graphic of the Week: Mapping Climate Change on Tatooine Over 110 Galactic Years

Wired News - Thu, 24/07/2014 - 9:06pm
Just because Luke Skywalker’s home planet of Tatooine is fictional doesn’t mean it’s immune to the effects of climate change. This map shows how, in the past 110 Galactic Standard Years, Tatooine has turned from a sprawling, desert wasteland into an even hotter sprawling, desert wasteland. It comes from Tatooine’s first Intergovernmental Report on Climate […]

Categories: Science

All You Need To Know About the 10% Brain Myth, in 60 Seconds

Wired News - Thu, 24/07/2014 - 9:06pm
The new Luc Besson movie Lucy, starring Scarlett Johansson, opens in theatres countrywide tomorrow. It’s based on the immortal myth that we use only 10 percent of our brains. Johansson’s character is injected with drugs that allow her to access 100 percent of her brain capacity. She subsequently gains the ability to learn Chinese in an […]

Categories: Science

A Warm-Feeling Wooden Keyboard (Video)

Slashdot - Thu, 24/07/2014 - 9:00pm
Plastic, plastic everywhere! Except on most surfaces of the Keyboardio ergonomic keyboard, which started as a 'scratch his itch' project by Jesse Vincent. According to his blurb on the Keyboardio site, Jesse 'has spent the last 20 years writing software like Request Tracker, K-9 Mail, and Perl. He types... a lot. He tried all the keyboards before finally making his own.' His objective was to make a keyboard he really liked. And he apparently has. This video was shot in June, and Jesse already has a new model prototype under way that Tim Lord says is a notable improvement on the June version he already liked. || Note that the Keyboardio is hackable and open source, so if you think you can improve it, go right ahead. (Alternate Video Link)

Read more of this story at Slashdot.

Categories: Science

Hubble Telescope Spies Off-Kilter Halo Around Distant Galaxy (Images) - Thu, 24/07/2014 - 8:53pm
The Hubble Space Telescope has created the most detailed map ever of the star-filled halo encircling Centarus A, the closest giant elliptical galaxy to our Milky Way galaxy. The map reveals a surprising amount of heavy metals along the halo's outer edge.
Categories: Science

ScummVM 1.7.0 Released

Slashdot - Thu, 24/07/2014 - 8:21pm
jones_supa (887896) writes It's been a while since a new ScummVM release, but version 1.7.0 is now here with many exciting features. New games supported are The Neverhood, Mortville Manor, Voyeur, Return to Ringworld and Chivalry is Not Dead. The Roland MT-32 emulator has been updated, there is an OpenGL backend, the GUI has seen improvements, AGOS engine is enhanced, tons of SCI bug fixes have been applied, and various other improvements can be found. This version also introduces support for the OUYA gaming console and brings improvements to some other more exotic platforms. Please read the release notes for an accurate description of the new version. SCUMM being the language/interpreter used by many classic adventure games.

Read more of this story at Slashdot.

Categories: Science

Laser Eye Surgery, Revisited 10 Years Later

Slashdot - Thu, 24/07/2014 - 7:42pm
gunner_von_diamond (3461783) happened upon Ask Slashdot: Experiences with Laser Eye Surgery from ten years ago, and asks: I was just reading a story on /. from 10 years ago about Lasik Eye Surgery. Personally, I've had Lasik done and loved every single part of the surgery. I went from wearing contacts/glasses every day to having 20/15 vision! In the older post, everyone seemed to be cautious about it, waiting for technical advances before having the surgery. Today, the surgery is fairly inexpensive [even for a programmer :) ], takes about 10-15 minutes, and I recovered from the surgery that same day. So my question is: what is holding everyone else back from freeing themselves from contacts and glasses?

Read more of this story at Slashdot.

Categories: Science