The "ClusterMeisters," a team of students from the Indiana University (IU) School of Informatics and Technische Universität Dresden (TUD), Germany were awarded first place in the SC08 Cluster Challenge, an international collegiate competition for leading-edge, energy-efficient high performance computing. The award was presented Thursday at SC08, the world's largest international conference for high performance computing, networking, storage and analysis, being held this week in Austin, Texas.

The goal of the annual Cluster Challenge competition is to showcase the ability of clusters and open-source software to solve interesting and important problems, and expose undergraduate students to state-of-the-art systems and the best and brightest of their peers. The team was supervised in the U.S. by Professor Andrew Lumsdaine and Torsten Hoefler of the IU Pervasive Technology Institute’s Open Systems Lab, and in Germany by Professor Wolfgang E. Nagel and Guido Juckeland of TUD's Center for Information Services and High Performance Computing (ZIH).

The U.S. and German students collaborated remotely for several weeks to prepare for the competition before finally meeting face to face in Austin. "Working with our partners from IU posed quite a challenge because the time to work together on a problem is very limited; it meant lots of night shifts" said Team captain Jupp Müller, a senior in Information System Technology at TUD. "This is also why the learning experience for all of us students was so great – since we always had to summarize daily work to our remote team members. But effort really paid off; it felt like all the pieces fell into place during the competition."

Working with corporate partners, IBM and Myricom, the IU-TUD team developed the most powerful system in the competition, while not exceeding the maximum electrical power allowed under challenge rules. IBM provided a new System X iDataplex server, designed to use up to 40% less power than similarly configured standard servers. This provided the team with a cluster that has a peak performance of over 1 teraFLOP. The latest Myricom networking equipment also used significantly less power than other high-performance networking technologies.

Also important to the ClusterMeister’s victory was application performance tuning, using state-of-the-art commercial compilers and advanced tools like Vampir, to analyze and improve the application behavior.

The ClusterMeisters achieved best overall performance for the High Performance Computing Challenge benchmark and five different scientific applications:

• The OpenFOAM (Open Field Operation and Manipulation) CFD Toolbox can simulate anything from complex fluid flows involving chemical reactions, turbulence and heat transfer, to solid dynamics, electromagnetics and the pricing of financial options. OpenFOAM, produced by OpenCFD Ltd, is freely available and open source, licensed under the GNU General Public License.
• The wave propagation program (WPP) developed by Lawrence Livermore National Laboratory implements substantial capabilities for 3-D seismic modeling. It is a parallel code for simulating time-dependent elastic and viscoelastic wave propagation, with some provisions for acoustic wave propagation.
•  POY4 is a flexible, multi-platform program for phylogenetic analysis of molecular and other data. An essential feature of POY is that it implements the concept of dynamic homology allowing optimization of unaligned sequences. POY4 can analyze entire chromosomes and genomes, taking into account large-scale genomic events (translocations, inversions, and duplications).
•  RAxML (Randomized Axelerated Maximum Likelihood) is a program for sequential and parallel Maximum Likelihood based inference of large phylogenetic trees. It was originally been derived from fastDNAml which, in turn, was derived from Joe Felsentein's dnaml, part of the Phylogeny Inference Package (PHYLIP).
•  The General Atomic and Molecular Electronic Structure System (GAMESS) is a general ab initio quantum chemistry package. It can be used to compute the properties of molecules and chemical reactions using a wide variety of theoretical models.

"I am truly impressed by the deep technical knowledge the students gathered in such a short time. This is their reward for many sleepless nights," said Matthias Mueller, Chief Technology Officer of ZIH, Technische Universität Dresden.

About Pervasive Technology Institute at Indiana University

Supported by a $15 million grant from the Lilly Endowment, Inc., the newly created Pervasive Technology Institute (PTI) at Indiana University combines intellectual talent from the former Pervasive Technology Labs, the IU School of Informatics, and University Information Technology Services. PTI leads IU in developing advanced information technology and informatics innovations and delivering their benefits to researchers, educators, students and society. PTI comprises three research centers, each composed of multiple labs: the Digital Science Center; the Data to Insight Center; and the Center for Applied Cybersecurity Research. IU is one of the oldest state universities in the Midwest and also one of the largest universities in the United States, with more than 110,000 students, faculty and staff on eight campuses. IU has a national reputation in the areas of information technology, informatics, and advanced networking.

About ZIH, Technische Universitaet Dresden

The Center for Information Services and High Performance Computing (ZIH) is a central scientific institution of the Dresden University of Technology. Beyond the support of local users ZIH cooperates with other HPC centers world-wide and is an established competence center for parallel computing and software tools. Since years, the widely used Vampir tool is developed by ZIH and provides a well-established and supportive performance analysis environment on HPC platforms. The vampir tool suite currently consists of the graphical user interface Vampir, the performance collection library Vampirtrace and the Open Trace Format (OTF).

About the Cluster Challenge

The cluster challenge is one of the annual challenges that is part of the Supercomputing Conference Series. More information is available at

http://sc08.supercomp.org/?pg=challenges.html