A team of students from the University of Utah’s School of Computing won a competition to build and run a small supercomputer cluster —a high-performance network used to perform intensive calculations for complex data sets such as weather forecasts or nuclear fusion.
Four national teams, including the U. students and a team from Skyline High School in Salt Lake City, were given identical components and two days to assemble and deploy a supercomputing cluster for a specific task. The competition, called the "LittleFe Challenge," was part of SC12, the annual international supercomputing conference held this year in Salt Lake City.
Utah computer science students Leif Andersen, Bruce Bolick, Ian King, Tom Robertson, Kathryn Rodgers and Tyler Sorenson were members of the winning team.
The competition involved what is known as the traveling salesman problem.
Three sets of cities and coordinates were given to each team. The students were then asked to use their small supercomputer to find the shortest route for a traveling salesman to take that would include a visit to each city once and a return to the city of origin for each data set using their cluster. The teams were judged on best score, visualization of their results and their knowledge of high-performance computing.
Brian Haymore, Martin Cuma and Wim Cardoen from the U’s Center for High Performance Computing were mentors to the team. Their faculty advisor was Mary Hall, a computer science professor.
Here's another reason to pay close attention to microbes: Current climate models probably overestimate the amount of carbon that will be released from soil into the atmosphere as global temperatures rise, according to research from the US Department of Energy's Lawrence Berkeley National Laboratory (Berkeley Lab).
The findings are from a new supercomputer model that explores the feedbacks between soil carbon and climate change. It's the first such model to include several physiologically realistic representations of how soil microbes break down organic matter, a process that annually unleashes about ten times as much carbon into the atmosphere as fossil fuel emissions. In contrast, today's models include a simplistic representation of microbial behavior.
The research is published Nov. 17 on the website of the journal Nature Climate Change.
Based on their results, the Berkeley Lab scientists recommend that future Earth system models include a more nuanced and dynamic depiction of how soil microbes go about the business of degrading organic matter and freeing up carbon.
This approach could help scientists more accurately predict what will happen to soil carbon as Earth's climate changes. These predictions are especially important in vulnerable regions like the Arctic, which is expected to warm considerably this century, and which holds a vast amount of carbon in the tundra.
"We know that microbes are the agents of change when it comes to decomposing organic matter. But the question is: How important is it to explicitly quantify complex microbial interactions in climate models?" says Jinyun Tang, a scientist in Berkeley Lab's Earth Sciences Division who conducted the research with fellow Berkeley Lab scientist William Riley.
"We found that it makes a big difference," Tang says. "We showed that warming temperatures would return less soil carbon to the atmosphere than current models predict."
Terrestrial ecosystems, such as the Arctic tundra and Amazon rainforest, contain a huge amount of carbon in organic matter such as decaying plant material. Thanks to soil microbes that break down organic matter, these ecosystems also contribute a huge amount of carbon to the atmosphere.
Because soil is such a major player in the carbon cycle, even a small change in the amount of carbon it releases can have a big affect on atmospheric carbon concentrations. This dynamic implies that climate models should represent soil-carbon processes as accurately as possible.
But here's the problem: Numerous empirical experiments have shown that the ways in which soil microbes decompose organic matter, and respond to changes in temperature, vary over time and from place to place. This variability is not captured in today's ecosystem models, however. Microbes are depicted statically. They respond instantaneously when they're perturbed, and then revert back as if nothing happened.
To better portray the variability of the microbial world, Tang and Riley developed a numerical model that quantifies the costs incurred by microbes to respire, grow, and consume energy.
Their model accounts for internal physiology, such as the production of enzymes that help microbes break down organic matter. It includes external processes, such as the competition for these enzymes once they're outside the microbe. Some enzymes adsorb onto mineral surfaces, which means they are not available to chew through organic matter. The model also includes competition between different microbial populations.
Together, these interactions-from enzymes to minerals to populations-represent microbial networks as ever-changing systems, much like what's observed in experiments.
The result? When the model was subjected to a 4 degrees Celsius change, it predicted more variable but weaker soil-carbon and climate feedbacks than current approaches.
"There's less carbon flux to the atmosphere in response to warming," says Riley. "Our representation is more complex, which has benefits in that it's likely more accurate. But it also has costs, in that the parameters used in the model need to be further studied and quantified."
Tang and Riley recommend more research be conducted on these microbial and mineral interactions. They also recommend that these features ultimately be included in next-generation Earth system models, such as the Department of Energy's Accelerated Climate Modeling for Energy, or ACME.
Spensa Technologies Inc., a precision agriculture company in Purdue Research Park, has launched a new Web and mobile enabled application to help growers and consultants more efficiently scout insects, weeds and disease, identify agronomic issues and nutrient deficiencies.
The new application, called "OpenScout," helps field and crop advisers to more easily identify and document the locations of insects, weeds, disease, nutrient deficiencies and general agronomic issues and assign a severity in a more precise manner so growers can strategically mitigate problems before they spread.
"Typically, a grower contracts with a consultant or other scouting service to track the crop throughout the growing season, and that is fine except that the information is commonly documented in a notebook and that is where mistakes can happen. This conventional technique often does not identify exactly where in the fields the problems are located or the severity of the problem," said Johnny Park, president and CEO of Spensa and a Purdue research assistant professor in electrical and computer engineering. "OpenScout records this data and it is automatically geo-tagged and time-stamped for easy retrieval and visualization."
OpenScout also tracks the scout's route through the field as a record when no observations are recorded.
Field problems can be documented with photos or text or the user can select from a crop level picklist. Once a field has been documented through Open Scout, an electronic heatmap illustrates high-level trends during the growing season across fields where problems are located so growers can more precisely make improvements.
Open Scout is the third technology developed by Spensa Technologies. Other technologies are an electronic insect-trapping system called Z-Trap that detects pests, and MyTraps.com, an online pest management program.
The Z-Trap and MyTraps.com products are currently used in five continents around the globe.
Spensa was named the top company in the 2013 BioCrossroads New Venture Competition. The company is located in the Purdue Research Park of West Lafayette and has received startup assistance through Purdue Foundry, an entrepreneurship and commercialization hub in Discovery Park's Burton D. Morgan Center for Entrepreneurship.
Untangling how cables coil
The world's fiber-optic network spans more than 550,000 miles of undersea cable that transmits e-mail, websites, and other packets of data between continents, all at the speed of light. A rip or tangle in any part of this network can significantly slow telecommunications around the world.
Now engineers at MIT, along with computer scientists at Columbia University, have developed a method that predicts the pattern of coils and tangles that a cable may form when deployed onto a rigid surface. The research combined laboratory experiments with custom-designed cables, computer-graphics technology used to animate hair in movies, and theoretical analyses.
In the lab, MIT engineers set up a desktop system to spool spaghetti-like cables onto a conveyor belt. They adjusted parameters such as speed of deployment and the speed of the belt, and observed how the cable coiled as it hit the surface.
At Columbia, computer scientists adapted a source code used for simulating animated hair and, incorporating the parameters of the MIT experiment, found that the simulation accurately predicted the coiling patterns seen in the lab.
The researchers say the coil-predicting method may help design better deployment strategies for fiber-optic cables to avoid the twisting and tangling that can lead to transmission glitches and data loss.
"We now have a set of design guidelines that allow you to tune certain parameters to achieve a particular pattern," says Pedro Reis, an associate professor of mechanical engineering and civil and environmental engineering at MIT. "We have a description that applies to many systems."
Reis and his colleagues publish their results this week in the Proceedings of the National Academy of Sciences. His co-authors are Khalid Jawed of MIT and Fang Da, Jungseock Joo, and Eitan Grinspun of Columbia University.
Shipping Up to Boston
Fiber-optic cables are typically deployed from a sailing vessel, which unfurls lengths of cable from a large spool. Depending on how the sailing speed of the boat relates to the speed of the spool, cable can be deposited on the seafloor in straight lines, or in meandering, coiling patterns.
"If the boat is sailing slower than the rate of the cable, then you're putting more cable down, which generates loops, coils, and tangles," Reis says. "That can lead to signal attenuation. But if the boat is traveling faster, then the cable can get taut and fracture, which is really bad news. So we wanted to understand what was underlying those patterns."
To do this, Reis set up a small-scale version of a cable-deploying system in his lab. He and his students fabricated filaments from silicone-based rubber, and rigged a spool to automatically reel out the wire onto a conveyor belt. They altered various parameters of the setup, including the speed of the belt and the spool.
The team used a digital video camera to record the filaments' motion as they hit the belt, and observed three main patterns: meandering waves, alternating loops, and repeated coils.
A Hollywood Makeover
To see if these patterns could be predicted in simulations, Reis teamed up with Grinspun, an expert in discrete differential geometry. Grinspun has applied sophisticated mathematical methods to simulating the movement of thin filaments such as hair and cloth — notoriously difficult features to animate realistically — for films including "The Hobbit" and Disney's "Tangled."
"The eye is very good at picking up what's physical and what's not," Grinspun says. "We want to capture the motion of hair and clothing in a realistic way, so a lot of algorithms we develop, we need to think about geometry."
Grinspun had previously upgraded a code he developed to simulate hair to model the flow of viscous fluids like honey. As honey is poured from a jar, it can resemble rope or thread, drizzling onto a surface in wavelike patterns. Reis wondered if the same code could be adopted to simulate the coiling of cables.
"We realized that I'm using geometry to scale up and down problems, and he's using geometry to speed up his codes, so we thought that we should port some of his algorithms into engineering, and test if these patterns can be predicted," Reis says.
At first, the collaborative effort produced mixed results: Patterns seen in experiments could not be replicated in simulations. The researchers eventually identified a key feature they were not originally factoring into the simulations: the natural curvature of the filament, which, when wound on a spool, retains a certain amount of curve as it's unwound. This initial mismatch between experiments and simulations motivated Reis to devise an experimental protocol to fabricate rods with customizable natural curvature.
With natural curvature now incorporated in the simulations and controlled in the lab, the researchers found that they were able to simulate the exact patterns observed in experiments. They then tuned the dimensions of various features in the simulation, and found they were able to predict the shape and amplitude of curves formed, based on several main factors: the speed of the wire deployed, the speed of the conveyor belt, the stiffness and diameter of the filament, and the size of its spool (a measure that determines a wire's natural curvature).
They also found, surprisingly, that the height from which a filament is deployed does not influence its coiling patterns — good news for ships that navigate choppy waters to deploy fiber-optic cables.
"This is important because, as a ship sails, the height of the ocean floor relative to the surface is changing all the time," Grinspun says. "We also know that how big you make the spools on the ship does matter. So we now have a map of how cables coil, and an understanding of what variables are important if you're trying to achieve certain patterns."
Going forward, Reis says that he and Grinspun may collaborate on other projects to understand and simulate the motion of thin filaments with features such as fluid drag and friction. For example, an understanding of such relationships from an engineering standpoint may improve the animation of phenomena such as hair blowing in the wind.
"I think what we now have is a bridge between these two fields, and we can start having traffic back and forth," Reis says.