Lafayette Online » supercomputer

Purdue highlights cloud computing tools at supercomputing conference

admin — Mon, 16 Nov 2009 18:39:50 +0000

WEST LAFAYETTE, Ind. — Facebook for scientists — but built to facilitate serious research rather than socializing — and an award-winning method for putting idle computers to work on scientific breakthroughs are Purdue-developed technologies in the spotlight at the SC09, the world’s largest high-performance computing conference.

Purdue's DiaGrid team includes, from left, Andy Howard, Phil Cheeseman, John Campbell, David Braun, Preston Smith and Carol Song. The team is posing with images from the scientific research enabled by DiaGrid, projected in a multiwalled virtual environment at ITaP's visualization facility. Campus Technology Magazine selected DiaGrid for a 2009 international Campus Technology Innovators Award. (Purdue University photo/Andrew Hancock)

Purdue is highlighting the HUBzero and DiaGrid technologies at the university’s booth at SC09, which opens today (Monday, Nov. 16) in Portland, Ore.

HUBzero is a soon-to-be open source software platform developed by Purdue for deploying and applying computational research tools, visualizing and analyzing results interactively and publishing them, all through a familiar Web browser. Built-in social networking features akin to Facebook create communities of researchers and educators in science, engineering, medicine and almost any field or subject matter.

DiaGrid works by pooling computers over the Purdue campus network and off campus via the Internet and fast research networks. Whenever machines in the pool are idle, such as at night or when their owners are at lunch, the system sends work to them. Campus Technology Magazine selected DiaGrid for a 2009 international Campus Technology Innovators Award.

Purdue has created an automated system to link the computers of SC09 participants to the pool during the conference. The Purdue booth includes a scoreboard to keep track of whose machines are running the most jobs.

The booth is designed to promote Purdue; Information Technology at Purdue (ITaP), the university’s central information technology organization; and the Rosen Center for Advanced Computing, ITaP’s research and discovery arm. ITaP technologists developed HUBzero and DiaGrid.

“DiaGrid and HUBzero are model technologies for enabling research that Purdue is making available to the world,” says John Campbell, associate vice president in charge of research computing for ITaP, who heads the Rosen Center. “As the premier conference for research computing, SC09 is a prime place to showcase these technologies.”

Purdue’s booth also will provide academic information to potential Purdue students and information to job seekers about positions with Purdue, ITaP and the Rosen Center. Nearly 10,000 people attended the conference in 2008.

Purdue has become a recognized leader in cyberinfrastructure with the development of HUBzero, which powers nanoHUB.org and many other Web-based “hubs” for research collaboration, says Michael McLennan, senior research scientist and hub technology architect at Purdue. NanoHUB is an international resource for nanotechnology theory, simulation and education with tens of thousands of users.

“Like no other platform, HUBzero can host interactive simulation tools. So, users aren’t just reading about research, they can experience it,” McLennan says. “HUBzero allows users to work together as they interact with content.”

Other hubs link researchers transforming laboratory discoveries into new medical treatments, and Purdue is now working in a consortium with Indiana and Clemson universities and the University of Wisconsin to advance the technology even further.

A hub will be at the center of the Network for Earthquake Engineering Simulation (NEES), a $105 Million National Science Foundation program announced in September, which is led by Purdue. Purdue electrical and computer engineering Professor Rudolf Eigenmann, co-principal investigator of NEES, will give a workshop titled “Cyberinfrastructure for Earthquake Engineering” at the Purdue booth.

McLennan will host two workshops on HUBzero and one about nanoHUB during the conference. Purdue scientist Mathieu Luisier will offer a workshop on using massive supercomputers to simulate nanoscale electronic devices for the next generation of electronics, a central focus of nanoHUB.

DiaGrid includes computers in student computer labs, offices, server rooms and supercomputing clusters and is the first multi-campus collaboration of its kind. Purdue’s partners in DiaGrid are IU, Indiana State University, the universities of Notre Dame, Louisville and Wisconsin, Purdue’s Calumet and North Central campuses, and Indiana University-Purdue University Fort Wayne.

Together, they now make nearly 30,000 processors available for research jobs ranging from understanding the Solar System’s formation to imaging the structure of viruses at near-atomic resolutions in an effort to develop new ways of battling viral illnesses, from swine flu and the common cold to West Nile virus and AIDS.

“The sheer size and ingenuity of the initiative, as well as the diversity of computing resources represented in the grid, really set the project apart,” Geoffrey Fletcher, editorial director of Campus Technology, said in announcing the Campus Technology Innovators Award for DiaGrid.

Purdue receives $105M to lead earthquake engineering network

admin — Fri, 11 Sep 2009 14:13:57 +0000

WEST LAFAYETTE, Ind. — Advancing research and education to reduce the devastation and loss of human life from earthquakes and tsunamis is the goal of a new center at Purdue University.

HUBzero is a new way for scientists and engineers to publish and share information. The latest hub will be used to study the causes and effects of earthquakes for the Network for Earthquake Engineering Simulation, or NEES. This hub joins others focused on topics such as nanotechnology, microelectromechanical systems, pharmaceutical products, cancer care, assistive technologies for people with disabilities, heat-transfer issues in engineering, and several others. New hubs are being created at a rate of about one per month. (Purdue University image/Michele Rund and Steve Tally)

The National Science Foundation awarded $105 million to a Purdue-led team to spearhead a center that will serve as headquarters for the operations of the George E. Brown, Jr. Network for Earthquake Engineering Simulation, or NEES. Submitted through the Cyber Center in Purdue’s Discovery Park, the grant spans five years and is the largest in the university’s history.

Purdue will connect 14 NEES research equipment sites and the earthquake engineering community through groundbreaking cyberinfrastructure, education and outreach efforts. Purdue’s center is expected to begin operations on Oct. 1, and will be housed in the university’s Discovery Learning Research Center in Discovery Park.

“I was delighted to learn that Purdue has the opportunity to lead this consortium of first-rate research universities,” said Purdue President France A. Córdova, who serves on the National Science Board, the governing board for the NSF, but was excused from deliberations on the award. “I’ve seen firsthand how devastating an earthquake can be not only to buildings, highways and the infrastructure of a city, but also to families, the community and people’s sense of security.

“Purdue’s depth of knowledge in earthquake engineering, innovative high-performance computing experts, education professionals and outstanding interdisciplinary research abilities allow the university to make great contributions to this area. The universities and institutions participating have had great individual successes, and we hope to bring them together to create a whole that is even greater than the sum of its parts.”

In the past decade, 124 major earthquakes have occurred throughout the world, according to the U.S. Geological Survey. Major earthquakes are generally accepted to register a magnitude in excess of 7 on the Richter scale and inflict serious damage, including the collapse of buildings and bridges, over a large area.

The organization estimates that earthquakes were responsible for 463,959 deaths in the past decade.

The Purdue-led NEEScomm Center, which stands for NEES Community and Communications, includes partners from the University of Washington at Seattle, University of Texas at Austin, University of Kansas at Lawrence, San Jose State University, the University of Florida at Gainsville, University of Michigan at Ann Arbor and Fermi National Accelerator Laboratory.

The center will help researchers share information and equipment to enable research and innovation in earthquake and tsunami loss reduction, create an educated work force in hazard mitigation, and conduct broader outreach and lifelong learning activities, said Julio Ramirez, the project’s principal investigator and a professor of civil engineering.

“Recent events have highlighted the importance of earthquake engineering for the United States and the world,” said Ramirez, who also will serve as the center’s director. “In China last year, tens of thousands of people perished as entire cities came down. The earthquake caused a tremendous financial burden in terms of rebuilding the civil infrastructure, but more than that, a high percentage of those who died were children – the future of a nation.”

Building codes and earthquake preparedness have improved in recent years, but even cities that lead in incorporating the latest safety features are at risk for serious damage, he said.

“Even in Los Angeles, a relatively young city in the global perspective, the building inventory includes structures that were built before buildings were engineered to mitigate earthquake hazard,” Ramirez said. “Earthquake design provisions have changed rapidly and substantially since 1980.”

Older cities and those in developing countries are likely to have the most buildings vulnerable to earthquakes and face the highest risk for damage and loss of lives, he said.

Forbes magazine cited a 2001 study by GeoHazards International that estimated the number of lives that would be lost if different cities experienced a magnitude 6 or higher earthquake. The study evaluated cities in Asia and the Americas. Kathmandu, Nepal, topped the list, followed by Istanbul, Turkey; Delhi, India; Quito, Ecuador; Manila, Philippines; and Islambad, Pakistan. All of the cities were estimated to face tens of thousands of fatalities if an earthquake struck.

The Purdue-based center will lead, manage, operate and maintain NEES. Through the NEES network, researchers from the United States and abroad conduct experiments and simulations of the ways buildings, bridges, utility systems and different materials perform during seismic events. Earthquake engineers will use this information to develop better and more cost-effective ways of reducing earthquake damage through improved materials, construction techniques and monitoring tools, Ramirez said.

The center’s education, outreach and training initiatives will use engineering education research to construct next-generation learning experiences that can be disseminated globally, he said. A team of engineering educators and technology specialists will develop the NEES Academy, a state-of-the-art virtual institution for cyber-enabled learning. The academy will help preK-12 teachers develop student interest in, and awareness of, science, mathematics, engineering and technology and will support undergraduate and graduate students engaging in research.

“The center will bring world-class education to any place in the world that has access to broadband Internet,” Ramirez said. “It will enable colleges and universities that don’t have the educational resources of Purdue to access the facilities, equipment, simulations and teaching tools of the participating institutions.”

Joy M. Pauschke, NEES program director at the National Science Foundation, said NEES is an integral part of the U.S. National Earthquake Hazards Reduction Program to support basic research to discover new knowledge, innovation and technologies for earthquake and tsunami loss reduction for the nation.

“During the first five years of operations, the unique NEES experimental capabilities and the extraordinarily strong user support from facility staff have enabled landmark testing and comprehensive experimental data capture for modeling seismic performance that was not possible before NEES,” Pauschke said. “Through Purdue’s leadership, the NEES experimental facilities, NEEShub and NEES Academy will provide world-class resources for earthquake engineering researchers, educators, students and practitioners not only in the United States, but globally.”

A cornerstone of the center is the development of information technology components that allow for new forms of collaboration and cooperation, said Rudolf Eigenmann, co-principal investigator and professor of electrical and computer engineering.

The NEEScomm Center will be a collaborative space and science hub where scientists and engineers can run scientific models and “what if?” scenarios. Hubs, which were first developed at Purdue, allow researchers to run models using a simple Web interface. The hub connects with supercomputing resources on the NSF’s TeraGrid and the national DiaGrid, which allow researchers to run their experiments without having to request time on a supercomputer.

Rudolf Eigenmann, professor of electrical and computer engineering, stands in front of screens displaying the prototype cyberinfrastructure for NEEScomm. The National Science Foundation awarded $105 million to a Purdue-led team to spearhead a center that will serve as headquarters for the operations of the George E. Brown, Jr. Network for Earthquake Engineering Simulation. Eigenmann is a co-principal investigator on the project. (Purdue News Service photo/Andrew Hancock)

“A critical and often difficult component of sharing data to advance research is making the findings of others easy to access and use,” Eigenmann said. “The HUBzero technology will allow for someone to simply go to a Web site and instantly be able to view data or run a simulation. It will eliminate the need to first download and install an application to view the data and then spend time downloading a large volume of data before beginning work.”

The cyberinfrastructure to be deployed by the NEEScomm Center is powered using HUBzero technology, which was originally created by researchers at Purdue University to support nanoHUB.org, a site for researchers studying nanotechnology. Currently there are nine other hubs online in science, engineering and medicine, and 12 more are expected to be online within a year, including the new NEEShub.

Ellen Rathje, professor of geotechnical engineering at the University of Texas at Austin and co-principal investigator of the project, said another key advantage of NEEShub is its data presentation capabilities.

“The new information technology created will enable research to take place that couldn’t be done before,” Rathje said. “Data will be presented in an easily searchable and usable format, like a virtual lab notebook, that will give experiments a longer life as researchers reuse existing data to run their own analysis and find new insights. It’s sort of like Facebook for scientists, but instead of posting vacation photos we’re posting research results.”

In addition to Ramirez and Eigenmann, Purdue University project co-principal investigators include Thomas Hacker, assistant professor of computer and information technology; Sean Brophy, assistant professor of engineering education; and Saurabh Bagchi, associate professor of electrical and computer engineering.

In addition, assistant professors of civil engineering Santiago Pujol and Ayhan Irfanoglu provide the team with earthquake engineering expertise in information technology matters of tele-operation and data management.

Project co-principal investigator Barbara Fossum, former managing director of Purdue’s Cyber Center, will serve as the center’s deputy director, and Dawn Weisman, former managing director of Purdue’s PRISM: National Nuclear Security Administration Center for Prediction of Reliability, Integrity and Survivability of Microsystems, will serve as information technology director.

Co-principal investigators Brophy and Thalia Anagnos from San Jose State University will co-lead the center’s education, outreach and training initiatives.

Fossum said the center will manage the operations of the network and engage the equipment sites to effectively carry out program initiatives.

“The United States has invested close to $300 million to commission and enable the equipment sites, and there are over 100 projects currently under way,” she said. “It is a resource for the nation and we want to help realize its full potential by sharing the unique resources and programs at each site across the network and with those interested in mitigating earthquake hazards throughout the world.”

The center will create more than 20 full-time positions and five graduate student positions at Purdue for Web application developers, software engineers, and research scientists and engineers.

The NEES equipment sites include Oregon State University; University of Nevada, Reno; University of California, Davis; University of California, Berkeley; University of California, Santa Barbara; University of California at Los Angeles; University of California at San Diego; University of Texas, Austin; University of Minnesota; University of Illinois, Urbana; University at Buffalo-SUNY; Cornell University; Rensselaer Polytechnic Institute; and Lehigh University.

NEES began development in 2000 and is a shared national network of 14 state-of-the-art earthquake engineering and tsunami experimental facilities at universities across the United States. It includes collaborative tools, a centralized data repository and earthquake simulation software.

Purdue builds Big Ten’s biggest computer, again

admin — Mon, 20 Jul 2009 15:53:56 +0000

WEST LAFAYETTE, Ind. — Purdue is acting as if building world-class supercomputers is the newest college sport.

For the second year in a row, Purdue will build what is expected to be the Big Ten’s largest campus computer, and as before, it will be running jobs by the end of the day.

Purdue University is installing the Big Ten's largest supercomputer, which will be built from more than 10,000 processors, or cores. To celebrate the new computer and the speed of its unique high-speed internal connections, Purdue produced a short video, Cores, which spoofs the Pixar hit Cars. (Purdue image /Michele Rund)

“Last year we unboxed the components for our Steele supercomputer in the morning and we were doing science in the afternoon,” says Gerry McCartney, chief information officer and vice president for information technology. “We expect to do the same thing with Coates, even though it is significantly larger.”

“Coates,” the new supercomputer, will be built from more than 10,000 computer cores, or processors, versus Steele’s 6,500 cores.

Coates is also expected to be the first internationally ranked academic supercomputer that is wired solely by superfast ten-gigabit network connections—allowing it to more easily handle the large amounts of data produced by research areas such as climate modeling and weather forecasting.

“Building supercomputers and other infrastructure needed for science and engineering is business as usual at Purdue,” McCartney says. “We have developed both a business model and an operational method that allows us to build world-class computers to meet the increasing demand from our researchers.”

On July 21, more than 200 information technology staff from Purdue will gather to construct the room-sized machine. They are expected to be joined by colleagues from the universities of Michigan and Iowa, as well Michigan State University and cross-state rival Indiana University, who will be observing and participating in the construction.

To generate excitement on Purdue’s campus and elsewhere, the IT staff created a parody movie trailer, “Cores,” which is a take off on the Pixar movie hit “Cars.” The video can be seen below.

Coates is expected to rank in the top 50 supercomputers worldwide when the next ranking is published in November. Supercomputers are ranked according to an agreed-upon benchmarking system, and the list is published twice a year at http://www.top500.org

The Big Ten’s largest campus computer is currently at the University of Minnesota’s Supercomputing Institute, which was ranked 59th in the June 2009 list, eclipsing Steele, which topped the list in its first benchmarking for the November 2008 Top 500 list. In that ranking, Purdue’s Steele was the Big Ten’s largest computer, ranked 105th in the world.

Supercomputing technology progresses rapidly, however, and six months later, in the June 2009 Top 500 list, Steele has moved from 105th to 196th.

“Even with Purdue’s international reputation as a leader in high performance computing, Coates isn’t being built for bragging rights,” McCartney says. “Top scientists and engineers require world-class resources in order to do their research, and with Coates we have a computer that is both powerful and capable of crunching massive data sets.”

Coates will be built with 1,280 HP dual quad-core computer nodes using AMD processors, and Cisco and Chelsio network equipment. It is expected to have a peak performance of 90 teraflops.

“Purdue University has deployed one of the world’s largest 10 GbE low latency, high performance computing infrastructures for scientific research, and we are honored that this strategic thought leader chose Cisco Nexus data center switches for a research facility of this magnitude,” said Soni Jiandani, vice president of marketing for Cisco’s Server Access and Virtualization Group. “Cisco is pleased to partner with Purdue to efficiently use computing resources and enable researchers to push the boundaries of science.“

Coates, like Steele, is being built as a “community cluster,” in which faculty on campus contributed research funds to fund the purchase, says John Campbell, associate vice president for Rosen Center for Advanced Computing at Purdue.

“Besides the cost savings from making a group purchase, the faculty can borrow computing cycles from other faculty when the other clusters are idle,” Campbell says. “This gives the researchers more flexibility, and we also have unused computing cycles we can offer to the National Science Foundation’s TeraGrid.”

The new cluster is being named for Clarence L. “Ben” Coates, head of Purdue’s School of Electrical Engineering (now Electrical and Computer Engineering) from 1973 to 1983. Coates retired in 1988 and died in 2000 at age 76. Coates was a driving force behind high performance computing at Purdue.

“Naming our research clusters after former Purdue IT leaders gives us a way to recognize the contributions of these great people,” McCartney says.

Game lets geeks compete to build virtual supercomputer

admin — Wed, 19 Nov 2008 18:34:07 +0000

WEST LAFAYETTE, Ind. — For those ready to get their geek on, Purdue University has created the computer game for you.

Rack-A-Node is an online video game that lets those über-geeks who love both science and technology try their hand at designing and operating a simulated research supercomputer.

Purdue University has created a game designed to teach how to build and operate a supercomputer. Called "Rack-A-Node," the game lets players design a supercomputer and then try to keep it operating as science jobs are submitted. (Purdue University image/Michele Rund)

“There’s a group of people who are into both science and computing,” says Kyle Bowen, informatics manager for Information Technology at Purdue. “The characters on the television show ‘Big Bang Theory’ would spend hours playing Rack-A-Node.”

The online game can be found at http://www.rcac.purdue.edu/rackanode/

Players build a cluster supercomputer using a variety of computing types to run science experiments. A player begins with a small supercomputer and receives science jobs to process. If these jobs are successful, the player receives funding needed to build an even bigger supercomputer.

The game requires the player to optimize the supercomputer to deal with waves of science jobs that are submitted.

“Like the game ‘rock, paper, scissors,’ certain tools perform better against certain challenges,” Bowen says. “In Rack-A-Node, the player has to optimize the supercomputer for the type of science being performed.”

For example, the game begins with a chemistry job that requires a lot of memory, then a climate-modeling job, which is a high throughput task that needs faster network communication. Later, a 3-D science animation-rendering job requires multiple nodes to process. The game also includes jobs from life sciences, pharmacy, physics and engineering.

“Supercomputing is not the most accessible of topics. It can be difficult to understand,” Bowen says. “We worked with a research scientist in Information Technology at Purdue to make sure the game is fairly realistic but still fun to play.”

The game was built to highlight Purdue’s student team participating in the Cluster Challenge at the SC ‘08 supercomputer conference on Nov. 15-21 in Austin, Texas. University teams compete in the challenge to see who can build the best supercomputer in a day.

In this year’s Cluster Challenge, the students will work with constraints on the amount of electricity they can use. Purdue’s team has partnered with SiCortex, a manufacturer of low-energy supercomputers. This summer Purdue was the first university to install a low-energy SiCortex supercomputer.

“Rack-A-Node is a game that captures the essence of the supercomputer challenge,” Bowen says.

Purdue’s Steele supercomputer makes list of world’s most powerful systems

admin — Wed, 19 Nov 2008 18:05:47 +0000

WEST LAFAYETTE, Ind. — Purdue’s Steele supercomputing cluster is among the most powerful high-performance computing systems in the world, according to rankings released Tuesday (Nov. 18) at the SuperComputing ‘08 conference in Austin, Texas.

The Top 500 Supercomputer Sites project has been ranking the 500 most powerful known computer systems twice a year since 1993 as a way of detecting and tracking trends in high-performance computing. Steele placed 105th on the latest list. Purdue ranked 319th in November 2007.

Steele ranked first among the Big Ten universities with systems on the list. Indiana’s Big Red cluster was at 148, and Minnesota had two entries that rank 268th and 356th. The Steele cluster is operated by Purdue’s Rosen Center for Advanced Computing, the research and discovery arm of Information Technology at Purdue, the university’s central information technology organization.

Gerry McCartney, Purdue’s vice president for information technology and chief information officer, said Steele’s showing was important not so much for where it puts Purdue on the Top 500 list as for the trend it indicates.

“The ranking of our new supercomputer, Steele, is just another indicator that Purdue is improving its position in the high-performance computing world. Our approach is drawing attention at the conference and of media,” McCartney said. “Of course, we don’t do this to see how high we can score on lists such as the Top 500. We do this to enable our scientists and engineers to stay at the forefront of discovery in crucial areas such as cancer research, global warming and the lack of affordable energy.”

Purdue is determined to continue enhancing the high-performance computing resources it provides for research and economic development purposes across the state, McCartney said.

Many people on the Purdue campus can take some of the credit for Steele’s placement on the list, announced at the premier international gathering for high-performance computing, networking, storage and analysis. Steele is a “community cluster,” funded by combining faculty grant and lab start-up funds and money from institutional sources.

Each “owner” gets a share of the computing power in the machine based on investment and the opportunity to tap more when they need from the shares of other users idle at the time.

“We built a top 500 machine by working collaboratively with the faculty,” said John Campbell, the associate vice president for information technology who heads the Rosen Center. “This machine is all about pulling together a diverse set of people, utilizing a variety of funding and sharing resources.”

Resources like Steele are integral to the research of Purdue faculty members who helped pay for the cluster, like Gerhard Klimeck, an electrical and computer engineering professor who models the next two or three generations of nanoscale electronic devices, allowing their properties to be understood long before they’re ever fabricated.

More than 250 staff members and volunteers assembled the cluster in a single Monday morning in May. Some of them even came from Purdue’s diehard in-state athletic rival Indiana, attracted by the idea of a high-tech barn raising to undertake a process that normally takes weeks.

Campbell noted that Steele recently averaged 87 percent owner utilization and more than 98 percent utilization overall.

That’s one reason the Rosen Center already is planning Purdue’s next community cluster, to be built in the spring of 2009. Faculty and campus organizations interested in participating in the new cluster, to be called Coates, can find more information online at http://www.rcac.purdue.edu/userinfo/resources/coates/

Tiny refrigerator taking shape to cool future computers

admin — Thu, 19 Jun 2008 15:54:25 +0000

WEST LAFAYETTE, Ind. — Researchers at Purdue University are developing a miniature refrigeration system small enough to fit inside laptops and personal computers, a cooling technology that would boost performance while shrinking the size of computers.

Researchers at Purdue are developing a miniature refrigeration system small enough to fit inside laptops and personal computers, a cooling technology that would boost performance while shrinking the size of computers. The researchers collect data using a myriad of sensors to precisely measure how a refrigerant boils and vaporizes inside tiny “microchannels” in a part of the refrigeration system called an evaporator. Data are needed to determine how to vary this boiling rate for maximum chip cooling. Eckhard Groll, at left, a professor of mechanical engineering, and Suresh Garimella, the R. Eugene and Susie E. Goodson Professor of Mechanical Engineering, discuss the microchannel data at the Ray W. Herrick Laboratories. (Purdue News Service photo/David Umberger)

Unlike conventional cooling systems, which use a fan to circulate air through finned devices called heat sinks attached to computer chips, miniature refrigeration would dramatically increase how much heat could be removed, said Suresh Garimella, the R. Eugene and Susie E. Goodson Professor of Mechanical Engineering.

The Purdue research focuses on learning how to design miniature components called compressors and evaporators, which are critical for refrigeration systems. The researchers developed an analytical model for designing tiny compressors that pump refrigerants using penny-size diaphragms and validated the model with experimental data. The elastic membranes are made of ultra-thin sheets of a plastic called polyimide and coated with an electrically conducting metallic layer. The metal layer allows the diaphragm to be moved back and forth to produce a pumping action using electrical charges, or “electrostatic diaphragm compression.”

In related research, the engineers are among the first to precisely measure how a refrigerant boils and vaporizes inside tiny “microchannels” in an evaporator and determine how to vary this boiling rate for maximum chip cooling.

The research is led by Garimella and Eckhard Groll, a professor of mechanical engineering.

“We feel we have a very good handle on this technology now, but there still are difficulties in implementing it in practical applications,” said Garimella, director of the Cooling Technologies Research Center based at Purdue. “One challenge is that it’s difficult to make a compressor really small that runs efficiently and reliably.”

Findings will be detailed in two papers being presented during the 12th International Refrigeration and Air Conditioning Conference and the 19th International Compressor Engineering Conference on July 14-17 at Purdue. The papers were written by doctoral students Stefan S. Bertsch and Abhijit A. Sathe, Groll and Garimella.

New types of cooling systems will be needed for future computer chips that will likely generate 10 times more heat than today’s microprocessors, especially in small “hot spots,” Garimella said.

Miniature refrigeration has a key advantage over other cooling technologies, Groll said.

“The best that all other cooling methods can achieve is to cool the chip down to ambient temperature, whereas refrigeration allows you to cool below surrounding temperatures,” he said.

The ability to cool below ambient temperature could result in smaller, more powerful computers and also could improve reliability by reducing long-term damage to chips caused by heating.

One complication is that the technology would require many diaphragms operating in parallel to pump a large enough volume of refrigerant for the cooling system.

“So you have an array of 50 or 100 tiny diaphragm compressors, and you can stack them,” Groll said.

The researchers conducted laboratory experiments with the diaphragms in Garimella’s Thermal Microsystems Lab, developed a computational model for designing the compressor and validated the model with data from the lab. Findings showed that it is feasible to design a prototype system small enough to fit in a laptop, Garimella said.

The model enables the engineers to optimize the design, determining how many diaphragms to use and how to stack them, either parallel to each other or in series.

“If you stack in one direction, you get more pressure rise, and if you stack in the other direction, you get more volume pumped,” Groll said.

Learning how to manufacture the devices at low cost is another major challenge, with industry requiring a cost of about $30 each.

“We can’t currently produce them at this price, but maybe in the future,” Groll said.

Another portion of the research focuses on learning precisely how refrigerant boils and turns into a vapor as it flows along microchannels thinner than a human hair. Such evaporators would be placed on top of computer chips.

Bertsch, the doctoral student who led work to set up experiments at the university’s Ray W. Herrick Laboratories, observed how refrigerant boils inside the channels and measured how much heat is transferred by this boiling refrigerant. He also created mathematical equations needed to properly design the miniature evaporators.

“This overall project represents the first comprehensive research to carefully obtain data showing what happens to heat transfer in arrays of microchannels for miniature refrigeration systems and how to design miniature compressors,” Garimella said. “Eventually, we will be able to design both the miniature compressors and evaporators.”

Some of the research was performed at the Birck Nanotechnology Center in Purdue’s Discovery Park.

The research is funded by the Purdue-based National Science Foundation Cooling Technologies Research Center, a consortium of corporations, university and government laboratories working to overcome heat-transfer obstacles in developing new, compact cooling technologies. Groll’s research is based at Herrick Laboratories.

Green supercomputer powers up at Purdue

admin — Mon, 16 Jun 2008 13:00:30 +0000

WEST LAFAYETTE, Ind. — Traditional supercomputers are like huge diesel trucks: They are powerful and can carry a big load.

But a new type of supercomputer installed this week at Purdue University, built by SiCortex, is like employing a fleet of thousands of bike messengers: Individually they don’t carry much freight, but they use less energy and are more effective in some situations.

Energy-efficient computing is becoming essential for science, says Purdue University’s Gerry McCartney, vice president for information technology and chief information officer, shown here standing in front of a SiCortex supercomputer that uses less power than traditional machines. McCartney says the energy costs of supercomputers is constraining science in some areas, prompting a need for new technologies. (Purdue News Service photo by David Umberger.)

The supercomputer is the first of what is expected to be many new technologies used at Purdue and elsewhere to lower the energy demands of scientific research.

Gerry McCartney, vice president for information technology and chief information officer at Purdue, says the supercomputer, a SiCortex 5832, uses 40 times less power than traditional supercomputers.

“The net energy savings is significant. We expect to see a reduction of 75 percent to 80 percent of the costs of the energy and the associated costs of the cooling in using this machine,” McCartney says. “But this is an experimental machine in the sense that we are just learning how to use it to make real scientific discoveries.”

SiCortex, based in Boston, is a new computing company that produces systems designed specifically for high performance computing used in research supercomputers. Argonne National Laboratory also has installed one of the energy efficient machines.

“Energy use has become one of the biggest challenges in conducting scientific research,” McCartney says. “It’s not just the power the computer itself uses, which can be significant, but also the air conditioning. Supercomputers are the prima donnas of the computing world. They’re like a star insisting on a special dressing room. These machines require special facilities called data centers where we pump in massive amounts of extra air conditioning or they literally self-destruct.

“The electricity needs of the supercomputers and the associated cooling can sometimes limit the amount of science that can be done.”

A recent research paper from the U.S. Department of Energy’s Lawrence Berkeley National Laboratory said that for a climate model to look at what is happening down to scales of one kilometer would require a supercomputer that would use 200 megawatts of power. That is the same amount of electricity used by a city with a population of 100,000.

“The energy issue is something we are going to have to solve in order to make significant scientific advances,” McCartney says.

The SiCortex 5832 is about the size of two refrigerators, and it has with DeLorean-style gull-wing doors, which gives it a “Back to the Future” vibe (although it is missing a flux capacitor).

Its processors draw just 600 milliwatts of power each, or about the same power requirements as a cell phone or small flashlight. By comparison, a standard supercomputer contains thousands of processors that require about 25 watts (25,000 milliwatts) of energy each.

The SiCortex computer uses a non-traditional architecture to achieve the power savings, and eliminates parts of the processor that aren’t needed for supercomputers that would otherwise expend energy. Also, because of the technology used, it is cheaper to buy than comparable supercomputers.

“With these advantages one might ask, ‘Why don’t you replace all of your supercomputers with this machine?’” McCartney says. “The reason is that this type of computer architecture works better for some kinds of science than others, and part of why we’re acquiring the SiCortex is so we can learn how to do more science with this type of computer.”

Research scientists in Purdue’s Office of Information Technology will explore what kind of computational tasks can work on the new class of machines and possibly even rewrite software so they can run common science applications.

Rudolf Eigenmann, professor of electrical and computer engineering and interim director of Purdue’s Computing Research Institute, says faculty researchers are already using the new machine.

“There are science applications that are already well adapted to this type of computing, such as research in chemistry and genetics, and even nano-electronics,” Eigenmann says. “We’ve put this computer to use from the first day, but we will also be looking for more areas in which we can use lower-power computing.”

Purdue scientists will be exploring new ways to conduct research so this type of computer machine can begin to replace other supercomputers on campus.

“At Purdue we have a team of talented research scientists in our central computing division who focus on improving scientific computing, so it makes sense for us to be among the first to look at this new technology to see how it can be used for discovery,” McCartney says.

Purdue supercomputer unboxed and built by lunchtime

admin — Fri, 09 May 2008 14:32:42 +0000

WEST LAFAYETTE, Ind. — Staff members at Purdue University had hoped to build the Big Ten’s largest campus supercomputer in just a day on Monday, May 5.

But it didn’t take that long — they were done by lunch.

Purdue computer technicians put the finishing touches on Steele, a new supercomputer that is among the largest in the world. Staff members at the university were challenged to build the supercomputer in a day, but finished the job by lunchtime. (Purdue News Service photo/David Umberger)

“The assembly was finished much faster than we expected, and by noon we were doing science,” says Gerry McCartney, vice president for information technology and chief information officer. “The staff was enthusiastic, the weather was great, and there were no problems installing the hardware or software. There is no cloud to accompany this silver lining.”

By 1 p.m. more than 500 of the 812 nodes that make up the supercomputer were already running 1,400 research jobs from across campus.

The supercomputer, which is named “Steele” for John Steele, former staff and faculty member, is made up of 812 Dell servers and is capable of performing 60 trillion operations per second. The supercomputer would rank in the top 40 of the current ranking of the world’s most powerful supercomputers, and is the largest supercomputer on a Big Ten campus that is not a part of a national center.

A time-lapse video of the supercomputer construction is available via YouTube: Supercomputer assembly at Purdue University

The first shift of workers was scheduled to begin unpacking boxes at 7 a.m., but many employees arrived at 6 a.m., eager to begin working. By 11 a.m. the supercomputer was essentially complete except for a few nodes that were intentionally held back to be installed at the noon dedication.

“We discovered that a build like this leverages the commodity nature of cluster computing, by using standard computing parts,” McCartney said. “By using commodity computer servers to build our supercomputer, we didn’t have to fly in engineers or hire specialized technicians. We were able to do it with our own IT staff in about four hours.”

Indiana University, Purdue’s rival on the athletic fields, surprised the Purdue IT staff by sending a crew of technicians to help build the machine.

Matt Link, director of research technology systems at IU, says he was pleased to be a part of the event.

“We often collaborate with people from Purdue on research proposals by videoconferencing, but we don’t routinely get the opportunity to work together in person,” Link said. “Our meeting today was enjoyable and will serve to strengthen future collaborations between IU and Purdue.”

The supercomputer was funded by Purdue faculty members who contributed research funds instead of purchasing equipment for their own laboratories.

Ashlie Martini, an assistant professor of mechanical engineering and one of the faculty who helped fund the project, will use the computer’s power to study friction at the molecular level. She watched the technicians install the nodes in the data center.

“The great thing about this approach is that almost everything was done for us,” Martini said. “This was very efficient. I have nothing but good things to say about today.”

Purdue installs Big Ten’s biggest campus computer in just one day

admin — Thu, 01 May 2008 20:26:26 +0000

WEST LAFAYETTE, Ind. — The largest supercomputer on a Big Ten campus will be installed at Purdue in a single-day, electronic “barn-raising.”

More than 200 employees will gather May 5 to help build the massive machine, which will be about the size of a semitrailer when installed. It will be the largest Big Ten supercomputer that is not part of a national center.

How much does 60 teraflops cost?

According to Wikipedia, the hardware cost of computing is $0.20 per gigaflop (October 2007 based on a Sony PS3). One gigaflop is 10^9, one teraflop is 10^12.

60 Tf = 60,000 Gf
60,000 Gf x $.20/Gf = $12,000

Purdue’s computer is being built in a single day to keep the university’s science and engineering researchers from facing a lengthy downtime, says Gerry McCartney, vice president for information technology and chief information officer.

“Our staff thought we were insane when we challenged them to build such a big computer in a single day,” McCartney says. “But now there’s real excitement to be a part of this.”

To generate interest on campus, the organizers created a spoof movie trailer called “Installation Day,” which is a take off of the movie “Independence Day.” The video can be seen on YouTube at http://www.youtube.com/watch?v=wVzThRN4QJI

Supercomputers are ranked by their performance in running a complex benchmarking system. The results of the tests are published twice each year at http://www.top500.org. Purdue’s new supercomputer would rank in the top 40 of the current Top 500 list, which was published in Nov. 2007.

The current campus leader in supercomputing in the Big Ten is Indiana University’s Big Red, which ranks 42nd in the world. (The National Center for Supercomputing Applications’ “Abe” cluster, which is based in Urbana, Ill. and operated by the University of Illinois, offers computing resources to researchers across the nation and is the largest supercomputer installed at a Big Ten university.)

The world’s largest supercomputer is BlueGene/L, which is located at Lawrence Livermore (Calif.) National Laboratory.

The Purdue supercomputer will consist of 812 Dell dual quad-core computer nodes and is predicted to have a peak performance of more than 60 teraflops, which means it could perform more than 60 trillion operations in one second.

A group of more than 25 university scientists and engineers pooled research grant funds to contribute to the purchase of the machine; less than 25 percent of the purchase is funded from the university’s IT budget.

“The community approach is a new and cost-effective way to fund cyberinfrastructure on campuses,” McCartney says. “This approach not only maximizes resources at Purdue, but researchers across the nation will benefit from the unused cycles which will be available on the TeraGrid.”

Rudolf Eigenmann, professor of electrical and computer engineering and interim director of Purdue’s Computing Research Institute, says the computer will be used for a wide variety of research.

“Faculty using this computer will be designing new drugs and materials, modeling weather patterns and the effects of global warming, engineering future aircraft, and making many more discoveries,” Eigenmann says. “High performance computing is an essential to conducting research and development, so having one of the world’s largest supercomputers here on campus will be a real benefit to our faculty.”

The three computing clusters that previously served university researchers were taken off-line and removed from the basement of Purdue’s Mathematics Building on April 28. A small number of the nodes for the new computer were then installed to serve as a “bridge” for continuing research until the new computer is fully installed.

An additional 154 nodes will be installed in Mann Hall. These nodes have already started practice computing runs for an international high-energy physics project that will begin this summer.

“We didn’t set out to acquire the largest supercomputer in the Big Ten. Our intent was to design a computer that would allow Purdue researchers to take the next step in discovery,” McCartney says.

The new computer will be named “Steele,” after John Steele, the former director of the Purdue University Computing Center, and a member of the Computer Science faculty, who retired in 2003.

“I appreciate that I can continue to be a part of high performance computing at Purdue and our efforts to remain at the forefront of this type of computing,” Steele says. “This machine will keep us on the high performance computing map.”

McCartney says Purdue plans to continue naming its major computers after faculty, staff or students who have made significant contributions to the university’s computing infrastructure.

“This isn’t the same as naming a building after someone; these machines have a lifecycle of about five years. But it is a way to salute members of the Purdue community who have worked so long and hard to help Purdue achieve the world-famous reputation it now enjoys,” McCartney says.

Purdue has a long history of leadership in information technology. In 1962, Purdue founded the first department of Computer Science. In 1967, Purdue became one of the first institutions to acquire a supercomputer, a Control Data Corp. 6500 (which had a performance of one-third of a megaflop). In 1982, Purdue.edu was the second URL registered for the Internet.

Purdue installs Big Ten’s biggest campus computer in just one day

admin — Thu, 01 May 2008 20:26:26 +0000

WEST LAFAYETTE, Ind. — The largest supercomputer on a Big Ten campus will be installed at Purdue in a single-day, electronic “barn-raising.”

How much does 60 teraflops cost?

According to Wikipedia, the hardware cost of computing is $0.20 per gigaflop (October 2007 based on a Sony PS3). One gigaflop is 10^9, one teraflop is 10^12.

60 Tf = 60,000 Gf
60,000 Gf x $.20/Gf = $12,000

“Our staff thought we were insane when we challenged them to build such a big computer in a single day,” McCartney says. “But now there’s real excitement to be a part of this.”

The world’s largest supercomputer is BlueGene/L, which is located at Lawrence Livermore (Calif.) National Laboratory.

Rudolf Eigenmann, professor of electrical and computer engineering and interim director of Purdue’s Computing Research Institute, says the computer will be used for a wide variety of research.

An additional 154 nodes will be installed in Mann Hall. These nodes have already started practice computing runs for an international high-energy physics project that will begin this summer.

The new computer will be named “Steele,” after John Steele, the former director of the Purdue University Computing Center, and a member of the Computer Science faculty, who retired in 2003.

McCartney says Purdue plans to continue naming its major computers after faculty, staff or students who have made significant contributions to the university’s computing infrastructure.

Lafayette Online » supercomputer

Purdue highlights cloud computing tools at supercomputing conference

In Related News:

Purdue receives $105M to lead earthquake engineering network

In Related News:

Purdue builds Big Ten’s biggest computer, again

In Related News:

Game lets geeks compete to build virtual supercomputer

In Related News:

Purdue’s Steele supercomputer makes list of world’s most powerful systems

In Related News:

Tiny refrigerator taking shape to cool future computers

In Related News:

Green supercomputer powers up at Purdue

In Related News:

Purdue supercomputer unboxed and built by lunchtime

In Related News:

Purdue installs Big Ten’s biggest campus computer in just one day

How much does 60 teraflops cost?

In Related News:

Purdue installs Big Ten’s biggest campus computer in just one day

How much does 60 teraflops cost?

In Related News: