Last week the House Committee on Science and Technology’s Subcommittee on Technology and Innovation held a hearing on the National Earthquake Hazards Reduction Program (NEHRP) in preparation for a new bill reauthorizing the program (the current authorization expires on Sep 30). From the Committee on Science and Technology’s website
“It is vital that we examine programs like this closely, since earthquakes and other natural threats can be devastating in their impact,” stated Subcommittee Chairman David Wu (D-OR). “For example, in the United States, wind and fire cause approximately $28 billion worth of damages and kill an average of 4,350 Americans each year. We can and must do a better job of hazards mitigation in order to protect our communities as much as possible from the devastation these disasters can cause.”
You care because parts of the earthquake prediction process are HPC jobs (see, for example, slide 16 of this presentation from 2007 to members of the House). The 2008 annual report also contains references to the program’s involvement with HPC, both direct and, in this example of the development of large scale dedicated resources, indirect
In 2007, the Southern California Earthquake Center (SCEC) began its third phase, a 5-year program supported primarily by the NSF and the USGS. SCEC is headquartered at the University of Southern California and unites 15 core institutions and 39 participating institutions in a “collaboratory” with a threefold mission: (1) gather data on earthquakes in southern California; (2) integrate these data and other information into a comprehensive, physics-based understanding of earthquake phenomena; and (3) communicate this understanding to the community at large as useful knowledge for reducing earthquake risk. In addition to core funding in 2007, the NSF provided support to SCEC to advance seismic hazard research using high-performance computing, with the aim of utilizing peta-scale computing facilities when they become available in the 2010– 2011 timeframe. (www.scec.org)
You also (should) care because the NEHRP has evidently done well, and might be a model to help fix the inadequacies of similar efforts to deal with other disasters.
Although the latest NEHRP authorization in 2004 also included a title creating the National Windstorm Impact Reduction Program (NWIRP), research for wind and other hazards is not yet produced the similar results.
“We’ve found that federal agencies currently have a stovepipe approach to hazards mitigation research activities,” said Wu. “Separate and distinct programs exist for earthquake, tsunami, fire, and wind threats, despite areas of commonality such as prediction research, emergency preparedness needs, and the potential for mitigation via enhanced construction codes. It is worth exploring whether a coordinated, comprehensive, and fully funded hazards mitigation program could be a more effective approach than the current stovepipe structure, where different hazards communities fight for their own funding priorities and lessons learned are less likely to be shared between those researching various threats.”
An interesting experiment in Japan on clustering together a couple NEC SX-9 vector machines over a long distance
The Cyberscience Center, Tohoku University, the Cybermedia Center, Osaka University, National Institute of Informatics (NII) and NEC Corporation jointly announced today the successful demonstration of one of the world’s fastest vector supercomputing environments by creating a single virtual system through the connection of two remotely located vector supercomputers on NAREGI (National Research Grid Initiative) middleware developed by NII.
The endpoints are at Tohuku U with a 16 node SX-9 and Osaka U with 10 nodes, and are about 500 miles apart (rough estimate from a quick peek at Google maps). They are connected via SINET3 (Science Information NETwork 3), Japan’s first 40 Gbps with 63 edge and 12 core nodes. The network brought Japan’s first 40 Gbps lines between Tokyo, Nagoya, and Osaka.
The software glue is the NAREGI middleware
NAREGI middleware enables large-scale computing resources at research and development centers scattered over a large area to be closely interconnected through high speed networks. These network connections can be viewed as a single massive virtual computer that efficiently implements large-scale parallel simulations, which were formerly difficult for individually isolated computer systems to carry out.
A new grid middleware component, the “GridVM for the SX Vector Computer,” was developed by enhancing the existing capabilities of the NAREGI middleware, such as job management, information provision and resource usage control. The enhanced GridVM maintains high compatibility with the local job scheduler (NQS) on the SX-9, which enables the efficient use of vector computing resources even in the grid environment. Moreover, it permits the co-existence of conventional (non-grid) jobs and grid jobs, allowing the computing center to provide a pioneering new cloud-computing service.
From the CCC blog, news that videos of the presentations given at their March Computing Research that Changed the World symposium are up and available for viewing. You can find them on the CCC’s YouTube channel, www.youtube.com/computingresearch. The talks include really interesting topics from speakers that will be familiar to many of you (Larry Smarr, Pat Hanrahan, Eric Brewer, and others).
NCSA’s Bill Kramer has finished volume 2 of his “Behind Blue Waters” video series (we covered volume 1 here)
Deputy Project Director Bill Kramer discusses NCSA and Illinois’ long-term collaboration with IBM. That work will improve the software that runs on the Blue Waters sustained-petascale supercomputer and helps make up the open-source environment that runs on other large-scale systems.
Also posted in HPC, HPC People
Our pal Joe Landman has a starring role in this article from GenomeWeb.com on the growing importance of GPUs in life science work
While there is never any shortage of vendors touting their hardware acceleration tool as the silver bullet for your computing bottlenecks, it’s important to keep in mind that it is not the hardware, but the algorithm and the data set that should inform your choice, along with cost and ease of use. The fact that GPU chipmaker NVIDIA has made porting code for GPUs easier for the average bench biologist with its CUDA software technology helps the argument for considering this breed of acceleration technology.
One of the things that strikes me as fundamentally different in life science work from the other computational fields I’ve been connected with is that the field actively encourages, through the open publication of genome and other data sets, anyone to join in the research. In some fields of life science study where you are concerned with pattern discrimination and matching, all you need is Perl, a laptop, and a network connection to play an active role in the research community.
This is a wonderful democratization of the research effort, and probably means that the speed of innovation and quality of invention in this field will dramatically outpace the other, higher barrier to entry fields that are dominated by the traditional large (rich) research institutions. Another application of my million monkeys coding postulate.
Of course the phrase “democratization” in computing also implies the phrase “I’m broke,” or at least “I’m not rich.” GPUs have disadvantages (programming), and they aren’t right for every job, but what they are is a cheap, low power source of compute. It makes sense that they would be enjoying increased attention in this field. And I think that because so many small scale teams and individual researchers are contributing here that there will probably be a noticeable network effort that helps GPU adoption — if Suzy Q is using them and having great results, then Johnny X will probably perceive them as a lower barrier path to entry and use them as well. I don’t perceive there is a similar network effect with either clusters or high end supers, probably because individuals don’t make these decisions, departments and institutions do, and they have different motivations other than just getting work done.
Also posted in Applied HPC, HPC
Intel announced yesterday that it will invest $12M US in a new computing research center in Europe
Intel Corporation is investing $12 million to create a new research center that will explore advanced graphics and visual computing technologies. Opening today, the Intel Visual Computing Institute is located at Saarland University in Saarbrücken, Germany. The investment, to be made over 5 years, represents Intel’s largest European university collaboration.
Applications studied by the center will include games (I know, not very HPC), medical imaging (getting warmer) and interactive exploration of large data models for scientific visualization (bingo).
The lab will conduct both basic and applied research in realistic, interactive computer graphics and natural user interfaces. By year’s end the institute will employ about a dozen researchers from such sources as Intel, Saarland University, Max Planck Institute for Informatics, Max Planck Institute for Software Systems and the German Research Center for Artificial Intelligence.
And the number of research staff is expected to grow to 60 over the next five years.
The Intel Visual Computing Institute will deliver more compelling visual computing applications through the development of new software designs and architectures, visual computing algorithms and parallel computing solutions. The institute will establish a feedback loop to Intel’s hardware design labs – including in Barcelona, Spain and Braunschweig, Germany — contributing to future visual computing hardware design. Current research contributions are expected to yield new software tools and hardware insights within just a few years.
HPC vendor SiCortex is talking about one of their customers, researchers and faculty at the Royal Military College of Canada, who are using their gear to design more fuel efficient airplanes
The aviation and shipping industries spend billions of dollars every year on fuel and account for 5 percent of total global carbon emissions. With fuel costs and global warming concerns on the rise, more energy-efficient aircraft are in demand. Researchers and faculty at the Royal Military College of Canada (RMC) are taking on this challenge, using a high-productivity computing (HPC) system from SiCortex to design more aerodynamic and efficient air vehicles. To achieve this, researchers are employing concepts like non-planar wing design and other surface configurations to positively impact aerodynamics, structural design, weight and performance.
The RMC’s mission goes beyond the important goal of improving fuel efficiency. By implementing a more energy-efficient, cross-functional design process, they are designing air vehicles that contribute to improved air transportation networks and reduced air traffic congestion. More broadly, the RMC is working in partnership with the National Defence and Canadian Forces to conduct highly-complex, advanced research to develop aircraft that better sustain damage during lengthy combat missions.
This just in via email, news that a research team from Argonne National Lab has been selected as one of the award winners at this year’s International Supercomputing Conference in Hamburg. From the abstract:
As high-end computing systems continue to grow in scale, the performance that applications can achieve on such large scale systems depends heavily on their ability to avoid explicitly synchronized communication with other processes in the system. Accordingly, several modern and legacy parallel programming models (such as MPI, UPC, Global Arrays) have provided many programming constructs that enable implicit communication using one-sided communication operations. While MPI is the most widely used communication model for scientific computing, the usage of one-sided communication is restricted; this is mainly owing to the inefficiencies in current MPI implementations that internally rely on synchronization between processes even during one-sided communication, thus losing the potential of such constructs.
In our previous work, we had utilized native one-sided communication primitives offered by high-speed networks such as InfiniBand (IB) to allow for true one-sided communication inMPI. In this paper, we extend this work to natively take advantage of one-sided atomic operations on cache-coherent multi-core/multi-processor architectures while still utilizing the benefits of networks such as IB. Specifically, we present a sophisticated hybrid design that uses locks that migrate between IB hardware atomics and multi-core CPU atomics to take advantage of both. We demonstrate the capability of our proposed design with a wide range of experiments illustrating its benefits in performance as well as its potential to avoid explicit synchronization.
This is the second year in succession that MCS Division researchers have won an outstanding paper award at the ISC. You can download the paper here [PDF].
The NSF.gov site is running a video interview with Kate Keahey of the Nimbus project, a cloud computing infrastructure developed at Argonne National Lab.
Nimbus is an example of such an adaptable system. Keahey and her team developed this open source cloud computing infrastructure to allow scientists working on data-intensive research projects to be able to use such virtual machines with a cloud provider. Nimbus also allows users to create multiple virtual machines to complete specific computational jobs that can be deployed throughout the cloud and still work in tandem with each other. This flexibility allows a user to configure a virtual machine and then connect it to resources on a cloud, regardless of who is providing the cloud.
Having this kind of flexibility and on-demand computing power is vital to projects that are extremely data-intensive, such as research efforts in experimental and theoretical physics. Nimbus has already been deployed successfully to support the STAR nuclear physics experiment at Brookhaven National Laboratory’s Relativistic Heavy-Ion Collider. When researchers there needed to turn the massive amounts of data they had generated into viable simulations for an international conference, they used Nimbus to create virtual machines that were run through commercial cloud computing providers.
The UK Met Office (the British government’s weather forecasting agency, founded in 1854) had some bad press in January of this year when the TimesOnline ran an article lambasting the organization for the carbon footprint of their new IBM supercomputer
For the Met Office the forecast is considerable embarrassment. It has spent £33m on a new supercomputer to calculate how climate change will affect Britain – only to find the new machine has a giant carbon footprint of its own.
“The new supercomputer, which will become operational later this year, will emit 14,400 tonnes of CO2 a year,” said Dave Britton, the Met Office’s chief press officer. This is equivalent to the CO2 emitted by 2,400 homes – generating an average of six tonnes each a year.
Now they are trying to get back out in front of that story with an emphasis on making energy considerations in the next upgrade more clearly decision drivers
The Met Office is planning to upgrade to its high performance computing systems in the next 18 months and is focusing on how to make those systems more efficient, according to the organisation’s head of IT services.
One of the techniques the Met Office has hit on is using direct current (DC) to power its servers rather than AC, to avoid the large losses of power during conversion from AC to DC, according to IT chief Steve Foreman, speaking at the Green IT ’09 conference in London, on Thursday.
…According to Foreman, the organisation is also looking at other ways to improve the efficiency of its high performance computing systems – used for weather modelling – such as increasing the temperature in its data centres.
That last bit is becoming quite popular; you’ll recall that Pete Beckman at ANL’s Leadership Computing Facility is doing the same thing with much success. Still, the UK Met is warning everyone (ahead of time for a change) that even though they are doing what they can, supercomputing still takes a lot of power
“Our supercomputers use something like 40 to 50 percent of our entire electricity usage in the organisation at the moment – that is about to go up to 80 percent,” he admitted. “Its going up because in order to provide more accurate weather information we need more computing power. We are getting more calculations per watt but the demand for calculations far exceeds the rate at which the suppliers are able to reduce the power power consumption.”
T-Platforms announced late last week that they’ve joined the ranks of vendors enthusiastic about NVIDIA’s Prefconfigured Tesla Cluster program
T-Platforms, the leading Russian supercomputer developer and manufacturer, and NVIDIA, the American manufacturer of graphic computing solutions, have become partners. Under the partnership agreement, T-Platforms is going to use NVIDIA Tesla solutions within its high performance computing systems and produce hybrid supercomputer complexes using Tesla C1060 graphics processors and Tesla S1070 servers.
…The partnership with NVIDIA has already yielded the first results. Quite recently, T-Platforms announced the release of a pre-configured hybrid T-Vision cluster based on T-Blade 1.1 blade servers developed by T-Platforms and computing nodes NVIDIA Tesla S1070. The cluster is available in two configurations: 15 and 30 computing nodes.
Yesterday IBM and the U of Texas announced that they were working together to use the World Community Grid in the fight against flu. Of course this is getting attention right now thanks to H1N1, commonly known as “swine flu.”
World Community Grid will run virtual chemistry experiments to determine which of the millions of small molecules can attach to the influenza virus and inhibit it from spreading. The computer predication can then be tested in the laboratory and clinic, which are the next phases in developing drugs for the marketplace. All of the results will be made freely available to other researchers studying influenza.
…World Community Grid is the largest public humanitarian grid in existence, with an 415,000-plus members linking more than one million computers. However, this represents only a fraction of the estimated one billion computers worldwide that could be used for medical breakthroughs. If more computers were contributing to the effort, the research could progress faster and more research projects, even those requiring the largest supercomputers, could be added.
Also posted in Applied HPC
This week visualization software provider CEI, maker of EnSight, announced the launch of EnSight CFD
This new CFD visualization product has an interface specifically tailored to the needs of CFD engineers, featuring improved readers for the analysis packages CFD engineers use most, including FLUENT and CFX. These enhanced readers automatically add a series of derived variables based on explicit dataset files. Therefore, quantities like radial velocity, pressure gradient, mach number, etc. are automatically available to the user, instead of having to manually create these quantities.
More info in the release.
From the BBC, news of work presented at the European Future Technologies meeting in Prague on “Blue Brain,” a research effort started in 2005 to reverse engineer mammalian brains from laboratory data
While many computer simulations have attempted to code in “brain-like” computation or to mimic parts of the nervous systems and brains of a variety of animals, the Blue Brain project was conceived to reverse-engineer mammal brains from real laboratory data and to build up a computer model down to the level of the molecules that make them up.
The first phase of the project is now complete; researchers have modeled the neocortical column – a unit of the mammalian brain known as the neocortex which is responsible for higher brain functions and thought.
Where are they at now? They’ve joined the brain model to a virtual animal in a virtual environment and are watching how the animal behaves in its environment
“It starts to learn things and starts to remember things. We can actually see when it retrieves a memory, and where they retrieved it from because we can trace back every activity of every molecule, every cell, every connection and see how the memory was formed.”
The next phase of the project will make use of a more advanced version of the IBM Blue Gene supercomputer that was used in the research to date.
Found at the NY Times, courtesy of the Computing Research Policy TumbleLog, an article about IBM’s research efforts to develop an application for its BlueGene platform that will let the computer compete with human contestants on Jeopardy! This is in the tradition of Deep Blue, the IBM SP that eventually beat champion chess player Gary Kasparov in a match in 1997.
But chess is a game of limits, with pieces that have clearly defined powers. “Jeopardy!” requires a program with the suppleness to weigh an almost infinite range of relationships and to make subtle comparisons and interpretations. The software must interact with humans on their own terms, and fast.
…The team is aiming not at a true thinking machine but at a new class of software that can “understand” human questions and respond to them correctly. Such a program would have enormous economic implications.
Happily, the computer will not be connected to the internet, but it will obviously have access to all the stuff it has synthesized prior to the contest.
I.B.M. will not reveal precisely how large the system’s internal database would be. The actual amount of information could be a significant fraction of the Web now indexed by Google, but artificial intelligence researchers said that having access to more information would not be the most significant key to improving the system’s performance.
Eric Nyberg, a computer scientist at Carnegie Mellon University, is collaborating with I.B.M. on research to devise computing systems capable of answering questions that are not limited to specific topics. The real difficulty, Dr. Nyberg said, is not searching a database but getting the computer to understand what it should be searching for.