The HPC Revolution 2005 is the premier international forum designed to foster interactions among people who are working with Linux clusters. From April 25- 28, 2005, users, researchers, vendors and developers of Linux clusters will gather in North Carolina at The Carolina Inn to learn about the latest hardware and software developments in the field.
Conference chair Dr. Douglas Pase is extremely excited for the conference. The gathering is on the University of North Carolina, Chapel Hill campus, close to the UNC computing facilities. Pase hopes attendees will be able to interact with other users and administrators of some of the largest computing clusters in the world.
Douglas Pase is the High Performance Computing Team Lead for eServer xSeries Performance Development and Analysis group at IBM. He has been an active participant in high performance computing since 1982, at NASA Ames Research Center, Cray Research, Inc., IBM and elsewhere. He currently studies all aspects of Linux cluster performance for scientific and technical work loads.
He has had to do quite a bit of work to organize the affair. He says everyone on the committee has worked hard to ensure an extraordinary conference, with more industry involvement to boost visibility and innovation.
Pase believes Linux clusters are shaping up to become the future of high- performance computing, as evidenced by a surge in cluster interest, as well as size.
“Just a short time ago, none of the systems on Jack Dongarra's list [of the world's fastest supercomputers] were Linux clusters,” says Pase. “Now, almost two-thirds of the systems are Linux clusters. This transition took place over the course of a few months. For this reason, we have brought in Mark Seager, from Lawrence Livermore National Laboratory, as one of our keynote speakers to talk about the direction of very large clusters.”
There will be three main keynotes taking place at the conference. The other speakers include Kelvin Droegemeier, Director, Center for Analysis and Prediction of Storms, University of Oklahoma and Thom Dunning, Director, National Center for Supercomputing Applications, University of Illinois, Urbana-Champaign.
Cluster adoption does not come without its drawbacks, however. Connecting scores of servers into one room can create an administrative nightmare, according to Pase. Pase hopes the LCI conference will forge novel ideas to overcome these hurdles.
“The tools are still under development that will help us turn a collection of cheap and very capable hardware [components] into a coherent and usable system,” says Pase. “It's a difficult problem. Our best hope of solving this problem is to get people together at conferences like the LCI, so we can sort out what works and what doesn't.”
Kelvin Droegemeier, a Regents' professor of meteorology and Weathernews (Inc.) Chair, directed a graduated NSF Science and Technology Center and wrote the proposal that started the first supercomputing center at the University of Oklahoma in the mid 1990s.
Droegemeier also helped co-found the current center as part of a $10M grant with Williams Energy Marketing and Trading Company. He says the center's main focus is to educate faculty and students about HPC use.
Droegemeier's keynote is entitled “Transforming the Sensing and Prediction of Intense Local Weather Through Dynamic Adaptation: People and Technologies Interacting with the Atmosphere.” He believes that HPC is the cornerstone of numerical weather prediction and has been since the first prediction experiment was conducted on the ENIAC in the late 1940s.
“HPC is transforming such prediction by allowing us to use much more sophisticated techniques — very computationally intensive techniques — to assimilate observations in models to yield much more accurate initial conditions. Further, large machines allow us to use increasingly finer grid meshes and thus resolve more detail,” says Droegemeier.
This April, Oklahoma University will launch the biggest prediction experiment ever performed. For this experiment, the community Weather Research and Forecast (WRF) model will be run daily over 2/3 of the continental US at 2km grid spacing.
To put the work in context with the conference, Droegemeier says that Linux clusters are being used everywhere in meteorological research and operations. OU's daily forecasts are run on a Linux cluster, for example.
In his talk, Droegemeier hopes to highlight the challenges that face meteorology as well as the enormous benefits to be gained through the use of adaptive systems that respond to the weather as it evolves. He says he will show results from experiments in which he has been able to predict the occurrence of individual thunderstorms a few hours in advance, and he'll describe the value of such information to airlines and other end users.
He hopes people come away with an appreciation of the value of HPC in meteorology, the pivotal role now played by Linux clusters, and the broader role of advanced cyberinfrastrucutre in ensuring the security and prosperity of our nation.
Thom Dunning will discuss what scientists and engineers must do to harness the power of high-end computing and information technologies to solve the most critical problems in science and engineering. He will examine the need for advances in the theoretical and mathematical sciences leading to computational models of ever increasing predictive power and fidelity, as well as the need for close collaboration among computational scientists, computer scientists, and applied mathematicians to translate these advances into scientific and engineering applications that can realize the full potential of high-end computers. Dunning also plans to address the task of educating a new generation of scientists and engineers to use computational modeling and simulation to address the challenging problems that they will face in the 21st Century.
For more information on the conference, please visit http://www.linuxclustersinstitute.org/Linux-HPC-Revolution/